GLM-5.2 Real-world Test - Code Generation Capabilities Rank Among the World's Top Tier

In the past, when domestically produced models were released, the best labels people would attach to various models were...open sourceFirst, it prioritizes cost-effectiveness.

But today is different. With the release of the Zhipu GLM-5.2 model, a deluge of news has flooded in.open sourceThe model is now on par with top-tier closed-source coding models. In terms of coding, the "Big Three" should become... GPT,Claude, it's intelligent.

GLM-5.2 is currently ranked first in Design Arena, with an Elo score of 1360.

Ranked #1 on BridgeBench BS (Anti-Nonsense Test) with a score of 100.0, and ranked #1 in Reasoning Ability with a score of 42.8.

GLM-5.2 also ranked 2nd in Code Arena: Frontend, compared to... Claude Opus 4.7 (Thinking) is 29 points ahead, only behind Fable 5!

However, as someone who isn't American, I'd like to ask you all... Claude Where are Fable5 making their fortune now? I can't seem to find a use for them!

The world's most powerful programming model was shut down by a ban, preventing its use, but GLM-5.2 offers capabilities of the same order of magnitude.open sourceIt was given to everyone.

However, I still plan to test GLM-5.2 myself. It's not that I don't trust these ranking lists, but I feel that testing it myself will help me determine whether the model is easy to use and whether it's suitable for my needs, since everyone's usage scenarios are different.

I plan to test it in two aspects: coding and daily tasks. Okay, enough talk, let the tests tell!

Case 1 1M Context Test

Prompt wordsBased on the document requirements, complete the design of the K-Sister's Canteen APP.

K-Sister's Cafeteria Takeout Ordering App Product Requirements Document (PRD)

Version: V1.0

Document Status: First Draft

Product Name: K-Sister's Canteen

Product Type: Food Delivery App

Target platforms: iOS, Android, H5, and mini-programs (expandable)

Target Phase: From MVP to Official Launch

Date of writing: 2026-06-17

1. Document Description

1.1 Document Purpose

This document defines the product positioning, business objectives, user roles, core processes, functional requirements, business rules, data metrics, backend management capabilities, non-functional requirements, acceptance criteria, and version planning for the "K Sister's Canteen" food delivery app.

The document is intended for roles such as product, design, R&D, testing, operations, merchant management, and delivery management, serving as a unified basis for subsequent prototype design, technical solution breakdown, R&D scheduling, test case writing, and online acceptance.

1.2 Product in one sentence

K-Sister's Canteen is a food delivery app targeting people who frequently eat on weekdays. It emphasizes "easy selection, quick ordering, on-time delivery, and easy repeat purchases," helping users complete their daily work meals, light meals, set meals, and drinks orders in the fewest steps.

Version 1.3 range

Version 1.0 focuses on the core closed loop of food delivery ordering, including user login, location selection, store browsing, menu selection, shopping cart, order confirmation, payment, merchant order acceptance, delivery fulfillment, order tracking, cancellation and refund, reviews, repeat purchases, and basic backend.

Version 1.0 does not include complex content communities, live streaming, group buying, membership tier systems, or other complex features.recommendAlgorithms, multi-person ordering, pre-made food e-commerce, and heavy marketing strategies prioritize ensuring a stable ordering process, accurate order status, traceable fulfillment, and manageable after-sales service.

2. Product Background

2.1 Market Background

Food delivery has become a frequent choice for weekday lunches, dinners, overtime meals, and quick snacks. Users' basic demands for food delivery platforms are not complicated: they want to find food quickly, have clear pricing, reliable delivery times, have solutions for problems, and be able to order again next time.fastRepeat purchase.

However, in actual use, many ordering experiences are disrupted by overly complex information feeds, too many marketing entry points, unclear menu information, unstable delivery promises, and hidden after-sales service entry points. Users don't want to browse the platform every time; often, they just want to decide "what to eat today" within one minute.

K-Sister's Canteen's opportunity lies not in becoming a large and comprehensive food delivery platform, but in first perfecting a high-frequency, stable, and controllable daily ordering scenario. This could include office parks, areas around schools, residential buildings, corporate meal subsidy scenarios, single-brand multi-store canteens, and chain light meal restaurants.

2.2 User Pain Points

Common pain points for users include:

1. High selection cost. The homepage is too cluttered with information, including events, advertisements, rankings, etc.recommendWhen everything is mixed together, users don't know where to begin.

2. The information about the dishes is unclear. There are significant differences between the pictures and the actual dishes, and the specifications, portions, spiciness, toppings, packaging costs, and delivery costs are not transparent.

3. Unstable experience during the lunch rush. Merchants are slow to accept orders, food preparation is slow, and riders have long waiting times, making it difficult for users to determine where they are stuck.

4. Inconvenient for repeat purchases. Frequently used meal deals and restaurants are not highlighted, and repeat purchases of past orders require reconfirming a bunch of options.

5. The after-sales process is cumbersome. When meals are missing, delivered to the wrong address, spilled, delivered late, or the rider cannot be contacted, users don't know what to do.fastdeal with.

6. High cost of understanding discounts. After discounts, coupons, delivery fees, and packaging fees are combined, users are unclear about why the actual price changes.

The pain points for merchants include:

1. During peak periods, orders are concentrated, which can easily lead to missed orders, incorrect orders, and asynchronous food preparation status.

2. The inventory and availability of dishes are not updated in a timely manner, and users only find out that the items are out of stock after placing an order.

3. There is a lack of standard procedures for handling refunds, menu changes, order reminders, and remarks.

4. The store's operating data is unclear, making it difficult to determine the best-selling dishes, repeat purchase dishes, reasons for negative reviews, and peak-hour pressure.

Operational pain points include:

1. Abnormal orders cannot be detected in a timely manner, and customer service passively accepts complaints.

2. The order taking, food preparation, and after-sales efficiency of different merchants cannot be quantified.

3. The promotional activities lack configuration and performance tracking.

4. Key metrics such as user retention, repeat purchases, refunds, and timeouts are scattered.

2.3 Product Opportunities

K-Sister's Cafeteria should focus on two aspects: "ordering efficiency" and "fulfillment certainty".

1. The homepage will not have a complex content flow; it will prioritize displaying nearby shops where orders can be placed, frequently ordered套餐 (set meals), and today's featured items.recommendAnd search.

2. Enhance the transparency of menu structure, specification selection, food status, and shopping cart amount on the store page.

3. The order status is refined, clearly displaying milestones such as merchant accepting the order, food preparation, rider picking up the food, delivery in progress, and delivery completed.

4. The repurchase entry point is placed at the front, allowing frequent users to complete the reorder with fewer steps.

5. The backend allows configuration of rules for stores, dishes, discounts, delivery, after-sales service, and handling of abnormal orders.

3. Product Positioning and Objectives

3.1 Product Positioning

K-Sister's Canteen is positioned as a high-frequency daily food ordering tool, not a general entertainment platform. The product experience should be clean, direct, and stable, minimizing unnecessary interruptions, and letting users know "what's available today," "how long will delivery take," "how much does it cost," and "who to contact if there are any problems."

Product keywords:

– High frequency

– Work lunch

– fastPlace an order

- On-time delivery

– Package deals

Easy to repurchase

– Traceable

– Clear after-sales service

3.2 Target Users

Core users:

1. Office workers. Their lunch and dinner needs are stable on weekdays, and they value delivery time, cost-effectiveness, and repeat purchase efficiency.

2. Students. Price-sensitive, prefer set meals, drinks, late-night snacks, and delivery to shared dormitories.

3. Community residents. They have a high demand for dinner, weekend meals, and occasional snacks, and value taste, vendor stability, and delivery coverage.

4. Corporate group meal/meal allowance users. Orders are placed in concentrated periods and require invoices, corporate accounts, fixed addresses, and bulk order capabilities. V1.0 only reserves expansion capabilities.

Secondary users:

1. Store clerk. Responsible for taking orders, preparing food, managing inventory, handling out-of-stock items, and confirming after-sales service.

2. Rider. Responsible for picking up and delivering meals, reporting any issues, and contacting users.

3. Platform Operations. Responsible for order monitoring, campaign configuration, merchant management, complaint handling, and data analysis.

3.3 Product Objectives

User experience goals:

1. From opening the app to completing their first payment, the core process for new users takes no more than 5 minutes.

2. Existing users can complete a repeat purchase through the "Order Again" option, ideally within 30 seconds.

3. Users can clearly see which fulfillment stage their current order is stuck at on the order details page.

4. Users can find the options to cancel an order, apply for a refund, contact customer service, or expedite an order within 3 clicks.

Business objectives:

1. Complete the closed loop of food delivery ordering, payment, order acceptance, food preparation, delivery, arrival, and evaluation.

2. Supports pilot launches for single stores and single business districts, with subsequent expansion to multiple stores.

3. Support basic discount configurations to improve first-order conversion and repeat purchases.

4. Establish basic data dashboards for orders, fulfillment, after-sales service, and merchant performance.

Technical objectives:

1. Order status is consistent and traceable, with no inconsistencies between the user end, merchant end, and backend status.

2. Key processes such as payment callbacks, refunds, and order cancellations are idempotent.

3. It should have basic resilience during peak periods, and order submissions and merchant order acceptance should not show significant anomalies due to concurrency.

4. Supports future expansion to multiple stores, multiple cities, corporate meal subsidies, membership programs, and more.intelligentrecommend.

4. Core Indicators

4.1 User Conversion Metrics

Homepage visitor count

- Number of visitors to the store page

- Add-to-cart rate for dishes

Shopping cart submission rate

Order payment success rate

First-order conversion rate

– Repurchase rate

- Average time spent on order placement

4.2 Performance Indicators

Average order processing time for merchants

Average food preparation time for businesses

Average delivery time per rider

Average delivery time per rider

- Average delivery time

- Overdue order rate

- Order cancellation rate

Out-of-stock order rate

– Abnormal order rate

4.3 Operating Indicators

– GMV

- Number of paid orders

Average order value

– Actual payment amount

– Subsidy Amount

– Delivery fee income

Packaging fee income

– Refund Amount

Merchant Sales Ranking

- Top-selling dishes

4.4 Experience Indicators

– User ratings

– Food rating

– Store rating

– Delivery rating

– Complaint rate

– After-sales processing time

– Refund approval rate

– Distribution of reasons for negative reviews

5. User Profiles and Scenarios

5.1 User Profile 1: White-collar workers who frequently have lunch

Name: Xiaolin

Age: 27

Scenario: Ordering lunch around 11:30 AM on a weekday.

Features: tight schedule, difficulty in choosing, frequent order of fixed meal packages

Demands:fastOrder placement, on-time delivery, stable prices, and repeat purchases are possible.

Typical behavior: After opening the app, browse frequently ordered stores or today's special offers, confirm the address, and place an order directly.

5.2 User Profile Two: Overtime Dinner Users

Name: A-Jie

Age: 31

Scenario: Ordering food while working overtime after 7 PM

Features: More concerned with delivery range and food preparation speed

Request: AbilityfastDo you know which stores are still open and how long the delivery is expected to take?

Typical behavior: Filter by "Still in operation", "Fastest delivery", and "Best-selling packages".

5.3 User Profile 3: Price-Sensitive Students

Name: Xiaoyu

Age: 20

Scene: Lunch, late-night snack, drinks

Features: Focus on discounts, spending thresholds, delivery fees, and minimum order amounts.

Demands: Clear actual price, easy-to-understand discounts

Typical behavior: Check the promotional packages and new customer coupons before deciding whether to place an order.

5.4 User Profile Four: Merchant Staff

Name: Sister Wang

Age: 38

Scenario: Orders are concentrated during the midday rush hour.

Features: Limited operation time, cannot frequently switch between complex pages.

Requests: Clear order notifications, clear remarks, and availability of out-of-stock items.fastdeal with

Typical behavior: Accept the order upon hearing the new order notification, prepare the food in order, and mark the food as ready for pickup after preparation.

5.5 User Profile Five: Platform Operation

Name: Mia

Age: 29

Scenario: Monitoring platform for orders and anomalies

Features: RequiredfastIssues such as timeouts, refunds, negative reviews, and merchant irregularities were discovered.

Requirements: A clear backend with filtering, export, and processing capabilities.

Typical actions: View the daily order dashboard and handle overdue orders and complaints in the exception pool.

6. Overall Business Process

6.1 Main Order Placement Process

1. The user opens K-Sister's Canteen.

2. The system obtains the location or the user selects the delivery address.

3. The homepage displays stores that offer delivery.recommendPackages and category entry points.

4. The user enters the store page.

5. Users select the dishes, specifications, and quantity, and add them to their shopping cart.

6. The user is taken to the order confirmation page.

7. The system verifies the delivery address, business status, inventory, minimum order price, discounts, and delivery fee.

8. Users select coupons, quantity of tableware, remarks, and payment method.

9. The user submits the order and completes the payment.

10. The system creates a payment order and waits for the payment callback.

11. After successful payment, the order will enter the waiting state for the merchant to accept the order.

12. Once the merchant receives the order, they begin preparing the food.

13. After the merchant prepares the food, it enters the waiting area for the rider to pick it up.

14. The rider picks up the food and begins the delivery process.

15. After the rider confirms delivery, the message will be displayed as "delivered".

16. User confirms receipt of meal or system confirmation.automaticFinish.

17. User reviews of orders.

6.2 Repeat Purchase Process

1. Users can access "My" or "Frequently Used" on the homepage.

2. The user selects historical orders.

3. Click "Order again".

4. The system verifies the store's business status, whether the dishes are available for sale, whether the specifications are available, and whether the inventory is sufficient.

5. If all items are available, restore the shopping cart and proceed to the order confirmation page.

6. If some dishes are unavailable, prompt the user to delete or replace them.

7. The user confirms and pays for the order.

6.3 Order Cancellation Process

1. Orders pending payment: Users can cancel them directly without a refund.

2. Payment made but merchant has not accepted the order: Users can cancel without penalty, and the system will handle the cancellation.automaticRefund.

3. The merchant has accepted the order but has not yet prepared the food: The user applies for cancellation, and the merchant will issue a refund after confirmation.

4. If the merchant has already prepared the food or the rider has already picked it up: cancellations without penalty are generally not supported and require customer service intervention.

5. Orders that have already been delivered: Cancellation is not supported; only after-sales service can be initiated.

6.4 Out-of-Stock Handling Process

1. The merchant discovers that the dish is out of stock before accepting the order.

2. Merchants can choose to refuse the order and fill in the reason, or initiate a request to change the menu/partial refund.

3. The user receives an out-of-stock notification.

4. Users can choose to accept replacements, remove out-of-stock items, or cancel their orders.

5. If the user fails to process the request within the timeout period, the system will proceed according to the configuration.automaticCancel or transfer to manual assistance.

6.5 Delivery Abnormality Process

Delivery anomalies include riders waiting too long at the store, users being unable to be contacted, incorrect addresses, damaged food, and delays caused by weather or traffic conditions.

Handling principles:

1. Riders must select the exception type and fill in a description.

2. The user client displays a simplified error message.

3. Records generated in the background exception pool.

4. The operations team can contact the user, rider, or merchant to resolve the issue.

5. The results of exception handling need to be synchronized to the order details.

7. Product Information Architecture

7.1 User-side information architecture

Primary Navigation:

- front page

- Order

- mine

Homepage Modules:

– Address bar

- search

– Category Entry

– Todayrecommend

- Frequently ordered dishes

– Nearby shops

– Promotional Activities

Order module:

– Current Orders

– Historical Orders

– Orders pending evaluation

– After-sales orders

My module:

– User Information

– Coupons

– Address Management

– Add to favorites

Customer Service Center

- bill

- set up

7.2 Merchant-side Information Architecture

Primary Navigation:

– Workbench

- Order

– Dishes

– Shop

- data

Workbench Module:

Operating Status

Today's orders

– Orders pending

– Preparing food

– Food ready to be served

– Abnormal alert

Menu Module:

– Menu

– Category Management

– Specification Management

Inventory Management

– Delisting

7.3 Backend Information Architecture

Primary Navigation:

– Data Dashboard

Order Management

– User Management

Merchant Management

– Rider Management

– Menu Management

– Discount Management

– After-sales management

System Configuration

8. User-side functional requirements

8.1 Login/Registration

8.1.1 Functional Description

Users log in via mobile phone verification code. First-time login.automaticRegister an account. V1.0 does not force users to set a password, does not allow binding of third-party accounts, and can reserve WeChat, Alipay, and Apple login extensions.

8.1.2 Requirements Details

1. The user enters their mobile phone number.

2. Click to get the verification code.

3. The system verifies the mobile phone number format.

4. After the verification code is successfully sent, the button will start a countdown.

5. The user enters the verification code and submits it.

6. After successful verification, you will be taken to the homepage.

7. If the user is a new user, the system will create a user account.

8. Users are not allowed to log in if they do not agree to the user agreement and privacy policy.

8.1.3 Exception Handling

- Incorrect phone number format: "Please enter the correct phone number".

– Verification code error: The message "Verification code is incorrect, please re-enter" appears.

– Verification code expired: The message "Verification code has expired, please obtain a new one" appears.

- Too many SMS messages sent: "Too many operations, please try again later."

8.2 Address and Location

8.2.1 Functional Description

Address is the foundation for food delivery fulfillment. The system needs to support this.automaticLocation, manual address selection, address addition, editing, deletion, and default address setting.

8.2.2 Requirements Details

1. When a user enters the app for the first time, the system requests location authorization.

2. After the user grants permission, the homepage displays nearby addresses of the current location.

3. When a user refuses location services, they can manually search for or add a new address.

4. The address field includes contact person, mobile phone number, gender and title, region, detailed address, house number, and tag.

5. Address tags support home, company, school, and custom.

6. Users can set a default address.

7. You must select a delivery address when placing your order.

8.2.3 Delivery Range Verification

Both when entering the store page and when submitting an order, it is necessary to verify whether the address is within the delivery range.

If not within the delivery area:

– The store card displays "Out of delivery range".

– Adding items to cart or completing the transaction is prohibited on the store page.

The order confirmation page prompts the user to change the address.

8.3 Homepage

8.3.1 Functional Positioning

The homepage is used to help usersfastOnce the ordering process begins, it does not handle complex content consumption functions. The homepage prioritizes resolving "What's nearby?" and "Today's menu".recommend"What?" "What do I usually eat?" "How quickly can it be delivered?"

8.3.2 Page Elements

1. Top address bar: Displays the current delivery address; click to switch.

2. Search box: Supports searching by store name, dish name, and keywords.

3. Category Entry Points: Work Meals, Light Meals, Noodles and Rice Dishes, Beverages, Late Night Snacks, Breakfast.

4. Todayrecommend: Packages or stores configured for operation.

5. Frequently ordered dishes: Displayed based on historical orders.

6. Nearby Shops List: Displayed in overall sorting.

7. Discount tags: Display new user coupons, discounts for purchases over a certain amount, delivery coupons, etc.

8.3.3 Store Card Fields

– Shop Name

– Shop profile picture or cover

– Rating

– Monthly sales

Average price per person

– Minimum order price

– Delivery fee

– Estimated delivery time

- distance

– Discount Tag

Operating Status

8.3.4 Sorting Rules

The default comprehensive sorting takes into account the following factors:

1. Is the business open for business?

2. Is delivery available to the current address?

3. Estimated delivery time.

4. Store rating.

5. Monthly sales.

6. User historical preferences.

7. OperationsrecommendWeights.

Version 1.0 allows for rule-based sorting without complex steps.Machine Learningrecommend.

8.4 Search

8.4.1 Functional Description

Users can search for shops, dishes, and keywords. Search results should clearly distinguish between shop results and dish results.

8.4.2 Search Entry

– Homepage search box

– Store page menu search

8.4.3 Search Results

Search results include:

1. Store list.

2. Menu.

3. No results page.

4. Historical search terms.

5. PopularSearch terms.

8.4.4 Handling No Results

When no search results are found, the following will be displayed:

– No relevant results found

– recommendUser changed keywords

– Show nearby best-selling stores or category entry points

8.5 Store Page

8.5.1 Functional Positioning

The store page is the core page for users' decision-making and ordering. Users should be able to clearly see whether the store is available for ordering, how long the delivery will take, what dishes are available, what the prices are, whether the dishes are selectable, and whether the shopping cart amount has reached the minimum for delivery.

8.5.2 Page Structure

1. Store header:

– Store Cover

– Shop Name

– Rating

– Monthly sales

– Announcement

Delivery time

– Minimum order price

– Delivery fee

2. Tag Page:

– Ordering food

- evaluate

Merchant Information

3. Sidebar for dish categories.

4. Menu list.

5. Bottom shopping cart.

8.5.3 Menu Information

The menu items field includes:

– Food photos

– Dish Name

– Dish Description

– Monthly sales

– Positive review rate

- price

– Original Price

- Specification

– Inventory Status

– recommendLabel

8.5.4 Food Condition

The status of the dishes includes:

– Normally available for sale

Sold out

– Removed from shelves

Supply outside the current time period

Unavailable dishes can be displayed but cannot be added to the purchase list; the button should be grayed out with an explanation.

8.6 Dish Specifications

8.6.1 Functional Description

Some dishes offer options for portion size, spiciness level, main course, toppings, and beverage temperature. When a user clicks to add a dish to their cart, a options panel will pop up if a specific option is available.

8.6.2 Specifications and Types

Common specifications include:

- Serving size: Small, Standard, Large.

– Spiciness level: Not spicy, Mild, Medium, Extra spicy.

Staple foods: rice, noodles, rice noodles.

– Add toppings: eggs, sliced meat, vegetables, cheese.

– Beverages: Hot, room temperature, iced.

8.6.3 Selection Rules

Specifications can be configured as single or multiple selections. If no mandatory specification is selected, it cannot be added to the cart. Specification surcharges must be added to the dish price in real-time.

8.7 Shopping Cart

8.7.1 Functional Description

The shopping cart is used to display the currently selected dishes, quantities, specifications, prices, discount information, and minimum order amount for delivery.

8.7.2 Requirements Details

1. Users can increase or decrease the number of dishes.

2. Users can empty their shopping cart.

3. Users can modify the specifications of selected dishes.

4. The total price of the dishes is calculated in real time in the shopping cart.

5. If the minimum order amount is not reached, a message will be displayed saying "You need X yuan more to order".

6. Once the minimum order amount is reached, the button will change to "Proceed to Checkout".

7. Only dishes from the same store are supported in the same order. V1.0 does not support cross-store shopping carts.

8.7.3 Price Display

The amount displayed in the shopping cart includes:

- Price of dishes

– Price increase for specifications

Packaging fee

– Estimated delivery cost

– Estimated Discount

– Expected actual payment

The full amount is subject to the order confirmation page.

8.8 Order Confirmation Page

8.8.1 Functional Description

The order confirmation page is the final confirmation page before payment. You need to confirm the delivery address, delivery time, menu details, discounts, remarks, tableware, fees, and payment method.

8.8.2 Page Fields

– Delivery address

– Contact person and phone number

– Estimated delivery time

– Shop Name

- Menu Details

Packaging fee

– Delivery fee

– Coupons

- Remark

– Number of cutlery

– Payment methods

– Fee details

– Submit Order Button

8.8.3 Pre-order verification

Orders need to be verified before submission:

1. Is the user logged in?

2. Is the address complete?

3. Is the address within the delivery range?

4. Is the store open for business?

5. Are the dishes still available for sale?

6. Is the inventory sufficient?

7. Does the amount meet the minimum order amount?

8. Is the coupon available?

9. Are payment methods available?

If the verification fails, the system must clearly indicate the reason for the failure and guide the user to make the necessary modifications.

8.9 Payment

8.9.1 Payment Methods

Version 1.0 supports WeChat Pay and Alipay, and allows for pre-paid balance payments and corporate meal allowance payments.

8.9.2 Payment Process

1. The user clicks "Submit Order".

2. The system creates an order, and the status is pending payment.

3. The system initiates the payment.

4. The user completes the payment.

5. Payment channel callback.

6. The system updates the order to "Pending Merchant Acceptance".

7. The user's device will display a payment success page or an order details page.

8.9.3 Payment Anomalies

– User cancels payment: The order remains in the pending payment status and can be paid again.

Payment failed: The reason for the failure will be displayed, and you can try to pay again.

- Payment successful but page not returned: The status displayed when the user enters the order page depends on the server status.

– Payment callback repetition: The server needs to handle this idempotently.

8.10 Order List

8.10.1 Function Description

The order list displays the user's current and historical orders, and supports viewing details, placing another order, leaving a review, requesting a refund, and contacting customer service.

8.10.2 Order Classification

- all

- in progress

– To be evaluated

– Refund/After-sales service

8.10.3 Order Card Fields

– Shop Name

Order Status

- Order time

– Menu Summary

– Actual payment amount

– Operation Buttons

The operation buttons change according to the status:

– Pending Payment: Proceed to Payment, Cancel Order

– Waiting for merchant to accept the order: Cancel the order, Contact the merchant

– Preparing food: expediting orders, contacting merchants

– During delivery: Contact the rider, check delivery status

– Completed: Place another order, leave a review, apply for after-sales service

– Cancelled: Place another order

8.11 Order Details

8.11.1 Function Description

The order details page displays the fulfillment progress and order information, and is the core page that users view after placing an order.

8.11.2 Page Module

1. Order Status Area:

– Current status

– Estimated delivery time

– Status Description

2. Timeline of contract fulfillment:

– Order submitted

The merchant has accepted the order.

– The restaurant is preparing the food.

– The rider has picked up the meal.

– In transit

– Delivered

3. Delivery Information:

– Rider's Name

– Rider's phone

– Current location, V1.0 may not require a real-time map, reserved for future use.

4. Merchant Information:

– Merchant's phone number

– Merchant Address

5. Menu details.

6. Detailed breakdown of expenses.

7. Order number.

8. After-sales service portal.

8.12 Evaluation

8.12.1 Function Description

Users can rate completed orders, including the overall order, the dishes, the delivery, and the merchant.

8.12.2 Evaluation Content

– Star rating

– Text review

– Image upload

– Anonymous review

– recommendLabel

8.12.3 Evaluation Rules

1. Completed orders can be rated.

2. Each order can only be reviewed once.

3. Once submitted, evaluations can be added within a certain period of time, but cannot be modified at will.

4. Information involving sensitive words, insults, or privacy will be subject to review.

8.13 After-sales service

8.13.1 After-sales service type

– Cancel order

– Request a refund

Partial refund

- Food shortage

– Damaged food

Wrong meal delivered

– Delivery timeout

– Meal not received

8.13.2 After-sales process

1. Reasons why users choose after-sales service.

2. The user fills in the instructions and uploads the credentials.

3. The system determines whether...automaticdeal with.

4. automaticIf the process fails or the amount is too high, it will be subject to manual review.

5. Merchant or platform operation and handling.

6. The user receives the processing result.

7. If the refund is approved, the money will be returned to the original payment account.

8.13.3 automaticRefund Policy

V1.0 can initially support a limited number of systems.automaticRefund scenarios:

- The user cancels the order before the merchant has accepted it.

– The merchant did not accept the order within the specified time.

– The store proactively rejected the order.

– Inventory verification failed after payment.

Other scenarios will be subject to manual review.

8.14 Coupons

8.14.1 Types of Offers

– Newcomer Voucher

- Discount coupon

- Delivery coupon

Store coupons

– Platform coupons

8.14.2 Usage Rules

1. New customer coupons are only valid for the first order.

2. Discount coupons require meeting a specified threshold.

3. Delivery vouchers can only be used to offset delivery fees.

4. Store coupons are only valid at designated stores.

5. Platform coupons are valid at all participating stores.

6. Only one discount coupon can be used per order by default.

7. Delivery coupons can be combined with discount coupons, depending on the backend configuration.

8.14.3 Display Rules

The order confirmation page should display available and unavailable coupons. Unavailable coupons should have a reason given, such as not meeting the minimum spending requirement, not applicable to this store, expired, or not eligible for combined promotions.

9. Merchant-side functional requirements

9.1 Merchant Login

Merchants log in using their account and password or mobile phone verification code. Accounts are created and permissions are assigned by the platform's backend.

Merchant account roles include:

– Shopkeeper

Store Manager

– Shop assistant

Different roles have different permissions. V1.0 can implement basic permissions first, allowing store staff to process orders and food inventory, and store owners to view data and manage store information.

9.2 Merchant Workbench

Workbench for peak periodsfastProcessing orders.

Content to be displayed:

– Current business status

- Number of orders today

– Today's sales

- Number of orders to be received

– Quantity in preparation

– Number of meals to be picked up by riders

– Abnormal order quantity

Core operations:

– Open for business

– Temporarily closed

– Order taking

Order rejection

– Mark the meal as served

– Contact User

– Contact the rider

9.3 Order Management

9.3.1 Order Status

Merchant-side order status includes:

– New orders

Order accepted

– Preparing food

– Food to be picked up

– Meal picked up

– Completed

– Cancelled

Refund in progress

9.3.2 New Order Notification

New orders should have a clear notification:

– Sound reminder

– Pop-up reminder

– Pin the list

– Unprocessed quantity subscript

If the merchant fails to process the request within the configured time, a timeout reminder will be generated in the backend.

9.3.3 Order Acceptance Rules

Merchants can accept or reject orders.

Order rejection requires selecting a reason:

- Dishes out of stock

- Closed

– Delivery error

– Order remarks could not be fulfilled

– Other reasons

After an order is rejected, it is cancelled and a refund is issued. The reason for the rejection is communicated to both the user and the backend.

9.4 Menu Management

9.4.1 Menu Field

– Dish Name

– Menu Categories

– Food photos

– Dish Description

– Price

– Original Price

- in stock

Packaging fee

- Specification

- whetherrecommend

– Available for purchase

– Sales Time

9.4.2 Food Preparation

– New dishes

– Edit dishes

– Delete dish

– Available for purchase

– Removed from shelves

– Set to sold out

– Restore inventory

– Adjust sorting

9.4.3 Inventory Rules

1. When inventory is 0, the dishesautomaticSold out.

2. Inventory is deducted after the user successfully places an order and makes payment.

3. Whether inventory is returned when an order is cancelled or refunded depends on the order status.

4. Merchants can manually set the "sold out today" option.

9.5 Store Management

Store information includes:

– Shop Name

– Store Logo

– Store Cover

– Store Announcement

Business Hours

– Delivery area

– Minimum order price

– Delivery Fee Rules

Packaging Fee Rules

– Contact number

– Store Address

Operating status includes:

- Open for business

– During break

– Orders suspended

- Closed

When order taking is suspended, users can browse the store and dishes, but cannot place an order.

9.6 Merchant Data

Basic operational data is displayed on the merchant's end:

- Number of orders today

Today's turnover

Today's refunds

– Bestselling dishes

- Negative review orders

Average order processing time

Average meal preparation time

Version 1.0 only provides basic display and does not perform complex business analysis.

10. Rider App Functionality Requirements

10.1 Rider Login

Rider accounts are created by the platform's backend, and riders log in via mobile phone verification code. Riders must have real-name information and contact details; in version V1.0, this can be entered by the backend.

10.2 Delivery Task

The rider's task list includes:

– Food to be picked up

– In transit

– Completed

– Abnormal Task

Task card display:

– Shop Name

– Store Address

– User address

– Estimated delivery time

Order number

– Contact the merchant

– Contact User

10.3 Status Operations

Riders can perform the following state operations:

1. Arrive at the store.

2. Meal picked up.

3. In transit.

4. Delivered.

5. Report anomalies.

Status changes need to be synchronized to the user's client, merchant's client, and backend.

10.4 Anomaly Reporting

Exception types include:

– The restaurant has not yet prepared the food.

- User cannot be contacted

– Incorrect address

– Damaged food

– Traffic reasons

– Weather

– Unable to enter the community or building

After a rider reports an anomaly, a record is generated in the backend anomaly pool, and operations can intervene to handle it.

11. Operations Backend Functional Requirements

11.1 Data Dashboard

The backend homepage displays the platform's overall operating data:

– Today's GMV

- Number of paid orders today

- Number of refunded orders today

– Today's active users

- Orders currently in progress

– Current abnormal orders

- Overdue order rate

– Average delivery time

Data can be filtered by date; V1.0 supports today, yesterday, the last 7 days, and the last 30 days.

11.2 Order Management

Order management supports viewing, filtering, exporting, and processing orders.

Filtering criteria:

Order Number

- User's mobile phone number

– Shop Name

– Rider Name

Order Status

– Payment Status

– After-sales status

- Order time

Order details display:

– User Information

Merchant Information

– Rider Information

- Menu Details

– Fee Details

– Payment Information

– State transition log

– After-sales records

Operation Log

Background execution is possible:

– Cancel order

– Initiate a refund

– Marking an exception

– Add a note

– Contact User

– Contact the merchant

– Contact the rider

11.3 User Management

User management fields:

– User ID

- Phone number

- Nick name

– Registration time

- Last order time

- Number of orders

– Payment Amount

– User Status

Background operations:

– View user details

– View user orders

– Disable user

– Unbanned users

– Add operational notes

Phone numbers must be anonymized; only authorized users can view the full phone number.

11.4 Merchant Management

Merchant management fields:

– Store ID

– Shop Name

– Contact Person

– Contact number

- address

Operating Status

– Review Status

- Number of orders

– Rating

Background operations:

– Add a store

– Edit Shop

– Review store

– Set business status

– Configure delivery area

– Configuration fee rules

– View operating data

11.5 Rider Management

Rider Management Fields:

– Rider ID

- Name

- Phone number

- state

– Number of delivery orders today

– Average delivery time

Number of complaints

Background operations:

– Add riders

– Editor Rider

– Enable/Disable

– Assign delivery area

– View delivery records

11.6 Discount Management

The backend supports creating coupons.

Coupon fields:

– Coupon Name

– Discount Type

– denomination

– Usage threshold

– Applicable stores

- Number of distributions

– Limited quantity per person

– Validity period

– Can it be stacked?

– Instructions for Use

Coupon status:

– Not started

- in progress

– Ended

– Discontinued

11.7 After-sales Management

After-sales management is used to handle refunds, complaints, and abnormal orders.

After-sales order fields:

– After-sales tracking number

Order number

– User

– Shop

– After-sales type

– Application amount

– Reason for application

– Certificate

– Current status

– Person in charge

– Processing results

After-sales status:

– Pending

– Merchant processing

– Platform processing

– Agreed

- Rejected

– Refund received

– Closed

The processing results must be recorded and cannot be deleted.

12. Order State Machine

12.1 Main State

Order master status includes:

1. Pending payment

2. Awaiting order acceptance from the merchant.

3. The restaurant is preparing the food.

4. Waiting for the rider to pick up the meal.

5. In transit

6. Delivered

7. Completed

8. Cancelled

9. Refund in progress

10. Refund has been issued.

12.2 State Transition Rules

Pending payment:

– The user has successfully paid and is now waiting for the merchant to accept the order.

– User canceled; access has been cancelled.

– Payment not made within the time limit, systemautomaticCancel.

Waiting for the merchant to accept the order:

– The merchant receives the order and begins preparing the food.

The merchant rejected the order; the page now shows "Cancelled and Refunded".

– If the user cancels, proceed to the "Cancelled and Refunded" section.

– If the merchant fails to accept the order within the specified time, the system will issue a notification orautomaticCancel.

The restaurant is preparing the food.

– The merchant marks the food as ready for pickup, and the order is ready for riders to collect.

- User cancellation requests require merchant confirmation.

Waiting for the rider to pick up the food:

– The rider picks up the food and begins delivery.

– When a rider reports an error, the process enters error handling, but the main state remains "awaiting pickup".

Delivery in progress:

– The rider confirms delivery and the page is now marked as delivered.

– The user reported that they had not received the product, and the process has been moved to after-sales service.

Delivered:

– User confirms receipt of meal or timeout.automaticCompleted, now in progress.

– The user applies for after-sales service and enters the after-sales process.

12.3 Status Log

Every state change must be recorded:

– Order ID

– Original state

– New Status

Operator type

Operator ID

Operation time

– Source of operation

- Remark

13. Fees and Settlement Rules

13.1 Composition of Order Amount

The amount payable on the order includes:

Menu item price + extra charge for size + packaging fee + delivery fee – discount amount = final payment amount

13.2 Packaging Fee

Packaging fees can be configured as follows:

1. Fixed packaging fee based on order.

2. Fee for individual packaging of each dish.

3. Packaging fees are added based on quantity.

Version 1.0 suggests supporting pricing based on food packaging fees, which can be configured in the backend.

13.3 Delivery Fee

Delivery fees can be configured based on distance, store, and order amount. Version 1.0 allows for simplified rules:

- Fixed delivery fee for the store.

- Free shipping on orders over a specified amount, as an extension to follow.

13.4 Refund Amount

Full refund:

The amount actually paid by the user will be refunded; whether or not coupons will be returned depends on the activity configuration.

Partial refund:

Refunds will be issued based on the actual amount paid for the dishes. If the discount involves a minimum purchase amount, it will need to be recalculated according to the backend rules or processed manually.

13.5 Merchant Settlement

Version 1.0 allows for initial data recording without establishing a complete financial settlement system. The following data needs to be recorded:

- Original price of the order

- Discount Amount

Platform subsidies

Merchants bear the discount

– Refund Amount

– Settlement Amount

14. Notifications and Messages

14.1 User Notification

User notification scenarios include:

Payment successful

The merchant has accepted the order.

– The restaurant has already prepared the food.

– The rider has picked up the meal.

– Rider is about to deliver

– Order has been delivered

Refund successful

– After-sales processing completed

Notification formats include in-app messages, push notifications, and SMS messages. Version 1.0 prioritizes in-app messages and push notifications.

14.2 Merchant Notice

Merchant notification scenarios include:

– New orders

– User cancels order

– User requests a refund

– Rider arrives at store

– Abnormal orders

New order notifications must be prominent and should support both sound and pop-up notifications.

14.3 Rider Notification

Rider notification scenarios include:

– New delivery task

– The restaurant has already prepared the food.

– User edits notes

– User contact request

– Delivery exception handling results

15. Permissions and Security

15.1 User Privacy

User phone numbers, addresses, and order records are considered sensitive information. The system must adhere to the principle of minimum visibility.

Require:

1. User mobile phone numbers are displayed in an anonymized manner by default.

2. Riders and merchants will only view user contact information during the necessary stages of order fulfillment.

3. Access to view the complete phone number in the backend requires specific permissions.

4. Address information must not be used for purposes other than order fulfillment.

5. Users can delete the address.

15.2 Backend Permissions

Backend roles include:

– Super Administrator

- operations

- customer service

- finance

Merchant Administrator

Permission example:

Customer service representatives can view orders and handle after-sales service, but cannot modify the discount rules.

– Operations can configure activities and handle abnormal orders.

– Finance personnel can view settlement data, but cannot modify order status.

– The super administrator can manage roles and permissions.

15.3 Operation Log

Critical backend operations must be logged:

– Modify order status

– Initiate a refund

– Modify merchant information

– Modify coupon

– Disable user

– Discontinued Merchant

– Modify delivery rules

16. Non-functional requirements

16.1 Performance Requirements

1. The homepage loading time should not exceed 2 seconds.

2. The response time for switching between menu items on the store page should not exceed 500ms.

3. The average response time for the order submission interface is no more than 1 second.

4. The order status will be updated within 3 seconds after the payment success callback.

5. During peak periods, the order status notification delay will not exceed 5 seconds.

16.2 Availability Requirements

1. Core order placement availability target: 99.9%.

2. Payment and order status must not be lost due to client exit.

3. After the merchant's network connection is restored, unprocessed orders need to be resynchronized.

4. Critical status information can be cached on the rider's device in the event of a weak network connection, and then retransmitted after the network connection is restored.

16.3 Compatibility Requirements

Mobile compatibility required:

– The last 3 major versions of iOS.

– Mainstream Android system versions.

– Mainstream screen sizes.

H5 or mini-program extensions need to be adapted to the WeChat environment.

16.4 Maintainability Requirements

1. The order status machine needs to be managed uniformly by the server.

2. Cost calculations must be performed uniformly on the server side; the client side only displays the results.

3. The discount rules should support backend configuration.

4. The rules for dishes, stores, delivery, and after-sales service should be modularized to facilitate future expansion.

17. Tracking Requirements

17.1 Homepage Tracking

– Homepage exposure

– Click to switch addresses

– Click the search box

– Click on categories

– Store card exposure

– Shop card click

– recommendPackage Click

17.2 Store Page Tracking

– Store page exposure

– Menu Revealed

– Add dishes to cart

– Specification pop-up window opens

– Specifications confirmed

– Shopping cart open

– Click to proceed to checkout

17.3 Order Tracking

Order confirmation page exposed

– Coupon Click

– Click to submit order

– Payment Initiation

Payment successful

Payment failed

– Click to cancel order

– Click for another order

17.4 After-sales tracking points

– Click here for after-sales service

– After-sales service type selection

– After-sales submission

– After-sales service successful

– After-sales service failure

18. Risks and Boundaries

18.1 Business Risks

1. During the lunch rush, orders are concentrated, putting significant pressure on merchants to receive and prepare meals.

2. Insufficient delivery capacity will directly affect the user experience.

3. Unclear discount rules can easily lead to price disputes.

4. Discrepancies between the restaurant's food photos and the actual dishes may result in negative reviews.

5. If after-sales review is too slow, it will amplify users' negative emotions.

18.2 Product Boundaries

Version 1.0 focuses on completing the order fulfillment loop, without pursuing complex marketing tactics. All new features should be prioritized based on whether they improve order placement efficiency, fulfillment stability, or after-sales processing efficiency.

18.3 Technological Risks

1. High consistency requirements for payment callback and order status.

2. Boundary issues can easily arise in inventory deduction and refund return.

3. The synchronization of status across multiple platforms, including the merchant's side, user's side, and rider's side, is complex.

4. Delays in push notifications during peak periods may affect contract fulfillment.

19. Acceptance Criteria

19.1 User-side acceptance

1. Users can log in using their mobile phone number and verification code.

2. Users can add, edit, and delete addresses.

3. Users can browse the stores and categories on the homepage.

4. Users can visit the store page to view the dishes.

5. Users can select the specifications and add the item to their shopping cart.

6. Users can submit orders and complete payments.

7. Users can view the order status.

8. Users can cancel eligible orders.

9. Users can apply for after-sales service.

10. Users can rate completed orders.

11. Users can place another order through their historical orders.

19.2 Merchant-side acceptance

1. Merchants can log in.

2. Merchants can switch their business status.

3. Merchants can accept or reject orders.

4. Merchants can mark the food preparation as complete.

5. Merchants can manage the listing and delisting of dishes and inventory.

6. Merchants can view today's orders and basic operating data.

19.3 Rider App Acceptance

1. Riders can log in.

2. Riders can view delivery tasks.

3. Riders can mark locations as arriving at the store, picking up food, or delivering it.

4. Riders can contact users and merchants.

5. Riders can report delivery abnormalities.

19.4 Back-end Acceptance

1. You can view the order list and order details in the backend.

2. The backend can manage users, merchants, and riders.

3. Coupons can be configured in the backend.

4. The back-end system can process after-sales orders.

5. Data dashboards can be viewed in the backend.

6. Key backend operations are logged.

20. Version Planning

20.1 V1.0 MVP

Objective: To complete the basic closed loop of food delivery ordering.

scope:

- User Login

– Address Management

- front page

– Store Page

- Dish specifications

Shopping cart

Order confirmation

– Payment

– Order Details

Merchants accepting orders

– Rider Delivery

– After-sales service

– Basic backend

20.2 V1.1 Experience Optimization

Objective: Improve repeat purchase and ordering efficiency.

scope:

- Frequently ordered dishes

– Add to favorites

– intelligentDefault address

– Improved rating tags

– Merchant food preparation time optimization prompt

– Couponsrecommend

20.3 V1.2 Operational Enhancements

Objective: To enhance the operational capabilities of events and merchants.

scope:

– Store promotion configuration

– Menu Ranking

– User Tier

– Business Operation Analysis

– Abnormal ordersautomaticWarning

– Analysis of reasons for after-sales service

20.4 V2.0 Expansion Capabilities

Objective: To support more complex scenarios.

scope:

- Ordering food for multiple people

– Corporate meal allowance

Invoice Management

Membership System

Group buying

- Scheduled delivery

– intelligentrecommend

Multi-city, multi-business district management

21. Prototype Page List

User side:

1. Splash screen

2. Login page

3. Locate the authorization page

4. Address list page

5. Added address page

6. Homepage

7. Search Page

8. Search Results Page

9. Store Page

10. Menu item specifications pop-up window

11. Shopping cart pop-up window

12. Order Confirmation Page

13. Payment Results Page

14. Order List Page

15. Order Details Page

16. Order Cancellation Page

17. After-sales application page

18. Review Page

19. My Page

20. Coupon Page

21. Customer Service Center Page

Merchant side:

1. Login page

2. Workbench

3. Order List

4. Order Details

5. Menu

6. Menu Editing

7. Store Setup

8. Data Page

Rider's App:

1. Login page

2. Task List

3. Task Details

4. Reporting anomalies

Backend:

1. Login page

2. Data Dashboard

3. Order Management

4. Order Details

5. User Management

6. Merchant Management

7. Rider Management

8. Coupon Management

9. After-sales management

10. System Configuration

22. Key Testing Points

22.1 Main Process Testing

– The entire process from logging in to placing an order and making payment.

– Merchant order acceptance process after successful payment.

– The process from when the merchant prepares the food to when the rider delivers it.

– User review process.

– Another order process.

22.2 Anomaly Testing

Payment failed.

Payment successful but client internet connection lost.

The merchant rejected the order.

– The merchant did not accept the order within the specified time.

– Insufficient food inventory.

– The address is outside the delivery area.

– Rider delivery error.

– The user requested a refund.

22.3 Amount Test

– Minimum order price calculation.

– Packaging cost calculation.

– Delivery fee calculation.

– Calculation of discount coupons.

– Delivery coupon calculation.

– Calculation of partial refund amount.

– Cancel the order and get a refund.

22.4 Permission Testing

Regular users cannot access the backend.

Merchants can only view orders from their own stores.

Riders can only view their own delivery assignments.

Customer service cannot modify the discount rules.

– Complete record of background operation logs.

23. Issues to be confirmed

1. Does V1.0 have to support both App and Mini Program simultaneously, or should we develop one platform first?

2. Is delivery handled by the platform's riders, by the merchant themselves, or both?

3. Does the payment need to support the balance or corporate meal allowance?

4. Is it necessary to integrate with the invoice system?

5. Was merchant settlement implemented in V1.0?

6. Does the map capability need to display real-time trajectory, or only the state transition?

7. Should the subsidies for new user coupons and discount activities be borne by the platform or the merchants?

8. After-sales serviceautomaticWhat is the maximum amount that can be refunded?

9. Is it necessary to support pre-orders and scheduled delivery?

10. Is corporate group meal service required?

24. Summary

The core of K-Sister's Canteen V1.0 is not to create a feature-packed food delivery platform, but to first ensure the smooth operation of the most important link in daily food ordering: [the user can...]fastCustomers can select meals, make payments clearly, and monitor status in real time; merchants can accept orders promptly and prepare meals accurately; riders can deliver smoothly; and the platform can detect anomalies and handle after-sales issues.

This version should be built around three keywords: efficiency, certainty, and traceability.

Efficiency is reflected in the homepage, store page, shopping cart, and repeat purchase path. Certainty is reflected in delivery range, inventory, price, order status, and estimated delivery time. Trackability is reflected in the order status machine, fulfillment logs, after-sales records, and backend operation logs.

If you have seen the abovePrompt wordsI knew it.Prompt wordsExtremely long, specifically designed for testing 1M contexts.

It's quite well-made; every little window is interactive.

The coverage is comprehensive, with a total of 19 display interfaces. Key pages such as the homepage, store page, specifications pop-up, order confirmation, order details, after-sales service, and reviews are all included.

The backend page is not just for show: it includes data dashboards, order management, and after-sales management.

All that's needed after adding interaction elements, status updates, real-world materials, and responsive design is to turn it into a complete app.

Case 2 3D Solar System

Let's start with a 3D project for our first test, a project that's been a staple in almost every model evaluation.

Prompt words:

Create an interactive 3D solar system page.

Require:

Use Three.js.

Planets revolve around the sun, and their orbits are visible.

After clicking on a planet, the side panel displays the planet's name, radius, distance, and description.

It supports play/pause, speed adjustment, view dragging, and scroll wheel zoom.

The mobile version has been changed to a top-bottom layout to prevent it from obscuring the main image.

The final result looks really good; it has all the necessary planets and offers comprehensive interactive features. It includes click, drag, zoom, pause, speed adjustment, and view reset.

It's clear that the model has a good foundation in Three.js, excellent page integration capabilities, and good interactive implementation capabilities.

The final result looks really good; it has all the necessary planets and the interactions are quite comprehensive. It includes click, drag, zoom, pause, speed adjustment, and view reset. Demonstration video: https://weixin.qq.com/sph/AobTs8a50

It's clear that the model has a good foundation in Three.js, excellent page integration capabilities, and good interactive implementation capabilities.

Case 3: Shooting Games

Next, we'll play a shooting game.

Prompt wordsPlease output a complete single-file HTML file and create a vertical scrolling shooter game similar to Raiden using Canvas. The player's aircraft can move and fire, enemy aircraft appear in batches and fire bullets; it should include collision detection, explosion effects, score, lives, levels, and a pause function; a boss will appear every 30 seconds; and mobile devices should have directional and shooting buttons.

I played for about ten minutes after finishing it, and it was really fun. It made up for the regret of not being allowed to play video games when I was a child. Demonstration video: https://weixin.qq.com/sph/AGFesU2uc

It has fighter jets, bosses, bullets, sound effects, and even screen shake. The gameplay structure is also complete, making it a full-fledged game. GLM-5.2 clearly understands the basic framework of vertical shooter games, managing to build a main loop, entity system, collision detection, mobile controls, and visual effects.

Case 4 bug fix

Next, let's test the bug fixing capability by fixing a Gantt chart HTML file. Gantt charts are a typical work scenario and can better demonstrate the front-end state design capabilities than ordinary tables.

Prompt words:

Below is a buggy single-file HTML file intended to create a sales trend chart. Please fix the code and output the corrected, complete HTML.

Don't just explain the problem; you must provide complete, directly executable code.

Original code:

 Repair requirements:

Fixed all data access issues that caused charts to fail to switch correctly.

Switching between Q1 and Q2 prevents the creation of multiple Chart instances, which can lead to memory leaks.

The page needs to be responsive, and the width should not overflow on mobile devices.

Add a KPI area to display total sales, total refunds, and net sales for the current quarter.

Add a button to switch between bar charts and line charts.

Add null state and error protection; if a non-existent quarter is passed in, the page should display a user-friendly message.

Finally, briefly indicate in the code comments which key bugs were fixed.

Results after repair: https://weixin.qq.com/sph/AS07Vq2uc

The code was thoroughly fixed. KPIs, chart type switching, responsiveness, and empty states were all added, indicating that the model understood the bug fix requirements.automaticThe user experience has been improved.

GLM has good front-end bug fixing capabilities: it can understand the original problem, fix core bugs, and proactively improve the product experience.

Case 5: Web Page Design

I think aesthetics are also a reflection of a model's capabilities. Let's try having GLM-5.2 generate an official website page.

Prompt words:

Please output a complete single-file HTML file, including HTML, CSS, and JavaScript, without relying on a backend. You can use CDNs such as GSAP, Three.js, or Lucide Icons, but do not use UI template libraries.

Subject: For a project called "LumaNote" AI Homepage of the official website for note-taking products.

Product Background:

LumaNote is designed for graduate students, product managers, consultants, and content creators. Core features include:automaticOrganize meeting recordings, turn long documents into structured notes, extract key points from multiple sources, generate traceable citations, and sync notes to Notion and Obsidian.

Page requirements:

The first screen must directly showcase the product; avoid vague, large titles. It needs a realistic product interface visual; you can use HTML/CSS to create a note-taking app interface, or Canvas/Three.js for interactive displays.

The first screen includes the product name, a clear one-sentence description, a main button, and secondary buttons.

The page should contain at least 5 complete sections: homepage, core workflow, key features, target audience, pricing options, and FAQ.

The core workflow should display "Import Data → automaticThe process of “organizing → generating reference notes → synchronization tools” cannot be summarized by simply listing the functionalities.

There should be at least 6 key features, each with an icon, title, and brief description.

The target audience should be 4 different user groups, and the copy should be written for different scenarios. The copy should not be repeated.

The pricing plan includesfreeThere should be three versions: Standard, Professional, and Team. Each version should have a price, be suitable for a specific user group, and include key features.

The FAQ should include at least 5 questions.

The page needs to have a top navigation bar and support smooth scrolling to the corresponding section by clicking.

Mobile devices must be adapted to avoid issues such as horizontal scrolling, text overflow, and button obstruction.

Design requirements:

The overall style should resemble a mature SaaS website: restrained, clean, and sophisticated.

Avoid using large areas of purple-blue gradients, floating colored orbs, emoji icons, and templated bento cards.

The corner radius of the card should not exceed 8px.

The main visual on the first screen cannot just be a decorative image; it must show how the product handles notes and citations.

Use a unified font hierarchy, white space, and color system.

Interactions should have details, such as highlighted navigation, expanded FAQ, button hover, workflow step switching or animation effects.

All copy should be in Chinese, using a tone similar to a genuine product website. Do not write it as... AI Promotional material for the product.

Output requirements:

Output only the complete HTML code.

All CSS and JS are written in the same file.

Do not explain the design concept.

The code should be able to be directly saved as .html and opened and run in a browser.

The official website page generated by GLM-5.2:

When I first opened it, I thought I had clicked on something else. AI Tools websiteAI  It feels unreal to see a website with such amazing aesthetics.

The warm paper-colored background, dark main buttons, thin borders, and low-saturation brown accent colors create a very comfortable feel. I've finally escaped the trap of just piling up gradient cards.

Case 6 Chinese Writing

Large ModelA person's Chinese writing skills are something that ordinary people value highly, especially since many office workers need them. AI Write some copy for yourself or something.

Prompt words:

Please write a 1200-1500 word Chinese article for a public WeChat account based on the following material.

theme:AI After a year with the tools, what are their truly useful and useless uses?

Background material:

A 12-person content team started using the system last year. AI Tools. The team primarily focuses on creating WeChat official account articles, short video scripts, and compiling industry materials.SimpleData analysis. After one year of use, weekly output increased from 3 articles to 5, and data processing time decreased by approximately 40%, but the initial draft revision rate did not decrease significantly. The team found that...AI It's best suited for data compilation, title selection, table cleaning, interview outlines, and draft structure; it's least suitable for writing direct opinions, judging industry trends, or making decisions for editors. Newcomers rely on it. AI Afterwards, articles are written more quickly, but the viewpoints often become superficial. (This is a common practice among experienced editors.) AI It's more like using an assistant; it saves time but doesn't abandon human judgment.

Writing requirements:

Create your own title; avoid clickbait.

Start by directly introducing the specific scene; avoid using phrases like "as..." AI Phrases like "development" and "in this era" are common.

Articles should contain personal judgment and should not be written as neutral reports.

It must be clearly stated:AI Where did it help, where did it not help, and why does the same tool have different effects in the hands of a novice and an experienced user?

Write at least 3 specific work scenarios.

Do not use terms such as "empowerment, reshaping, ecosystem, closed loop, underlying logic, paradigm, or dimensional reduction attack".

Do not use the "not...but..." structure.

Don't write every paragraph as short, catchy sentences.

The conclusion offers a specific suggestion: do not elevate the discussion to a higher level.

The tone is natural, like a debriefing written by a real content team leader.

Pretty fast, they were out in less than two minutes!

The article reads smoothly overall, and the beginning is quite engaging.

The comparison between newcomers and veterans is a big plus, giving a sense of real management experience and making it more memorable than simply talking about the advantages and disadvantages of tools.

If the maximum score is 100, I would give GLM-5.2 writing ability an 85.

Case 7 instruction follows

Many models often misunderstand our instructions; let's see if GLM-5.2 has this problem.

Prompt words:

Please process the text according to the following rules. Rules are listed in descending order of priority; in case of conflict, only the higher-priority rule will prevail.

Rule A: The final answer can only output 4 bullet points.

Rule B: Each bullet point must contain fewer than 18 Chinese characters.

Rule C: Numbers in the original text must be retained.

Rule D: Do not use the words "improve", "optimize", or "create".

original:

The system is expected to go live in 30 days, with the goal of improving customer service response speed, optimizing work order allocation process, creating a unified service portal, and reducing the proportion of manual processing to 40%.

Very good. It meets the requirements of 4 bullet points, each with fewer than 18 Chinese characters, retaining numbers, and avoiding prohibited words.

Case 8: A classic trap question

Prompt wordsI need to get my car washed. My house is 50 meters away from the car wash. Should I drive there or walk?

Thankfully, we didn't fall into a trap. They even kindly suggested that we come back first and pick up the car after washing it.

Case 9: PPT Creation

Prompt words:

Prompt wordsPlease create a PowerPoint presentation of no more than 8 pages based on the following material. The subject is "AI "Tool Implementation Plan in Content Teams". Requirements: Include a cover page, current issues, goals, process design, job responsibilities, risk control, pilot plan, and closing page. Use a solid business style, no cartoon illustrations. Each page should contain only necessary text, avoiding excessive paragraphs. Include at least one flowchart and one table. Generate an editable PPTX file and export a preview to check the layout. Materials are as follows:

Project Background:

The company is a SaaS service provider serving B2B clients, primarily in the manufacturing, retail, education, and professional services industries. The content team is responsible for WeChat official account articles, client case studies, product white papers, short video scripts, sales materials, and event promotional materials.

Over the past year, content demand has increased significantly. The marketing department hopes to increase monthly content production from the current 18 articles to 30 articles, including:

Eight in-depth articles on the WeChat official account.

Four client case studies.

12 short video scripts.

Four sets of industry data were compiled.

Two copies of sales support materials.

The current team has already tried using AI Tools are available, but the process largely depends on personal habits; there's no standardized workflow. Some editors use... AI Some editors rarely do the work of compiling data and preparing title options. Output quality is inconsistent, and review standards are not uniform.

Team size and roles:

The content team consists of 12 people.

One content manager is responsible for topic selection, final review, and cross-departmental communication.

Three senior editors are responsible for in-depth articles, client case studies, and white papers.

Three junior editors are responsible for data organization, drafting, and rewriting short content.

Two short video directors are needed, responsible for scripts, storyboards, and narration.

One designer is needed to design the cover image, infographics, and PPT visuals.

One person is responsible for operations, including publishing on the official WeChat account, data collection, and distribution to communities.

One data analyst is needed to analyze content effectiveness and generate monthly reports.

Current main problems:

Data processing takes a long time.

On average, it takes 6 to 8 hours to collect, extract, and verify information for an in-depth article.

The quality of the first draft fluctuated greatly.

First drafts written by junior editors often have a complete structure, but the viewpoints are not clear enough, resulting in a revision rate of about 38%.

Titles and abstracts rely on experience.

Different editors have very different writing styles, and the title of the same article often needs to be revised 4 to 6 times.

Customer case production is slow.

After the interview recordings are transcribed, the key points need to be manually sorted out. On average, it takes 7 days from interview to final draft for one case study.

Insufficient content reuse.

A white paper is rarely broken down into short video scripts, sales pitches, or short articles for social media, resulting in low utilization of the material.

Risk control is unstable.

AI Occasionally, the generated content may contain factual errors, unclear citations, exaggerated product effects, or overly marketing language.

Existing data:

Current monthly output: 18 articles.

Target monthly output: 30 articles.

Average production time for in-depth articles: 3.5 days.

Average production time for client case studies: 7 days.

The initial draft revision rate was 38%.

Average time spent on data processing: 6 to 8 hours per article.

The average number of title revisions is 4 to 6.

Average completion rate within 7 days of content publication: 42%.

The sales team needs ad-hoc data approximately 15 times per month.

Implementation goals:

Increase monthly content output to 30 articles within 3 months.

Reduce the time spent compiling in-depth articles and materials by 40%.

We shortened the client case production cycle from 7 days to 4 days.

The initial draft revision rate was reduced from 38% to below 25%.

Establish a unified AI Use guidelines and audit checklists.

Each key piece of content should be reused in at least three formats, such as WeChat official account articles, short video scripts, and sales pitches.

List of available tools:

ChatGPT or similarLarge ModelUsed for data summaries, outlines, title options, and rewriting of initial drafts.

Perplexity or similar search tools: used for retrieving publicly available information and tracing sources.

Lark Docs: Used for collaboration, version control, and approval workflows.

Notion, or a knowledge base tool: used to store industry information, customer case studies,Prompt wordstemplate.

CapCut or Jianying: Used for splitting short video scripts and creating initial subtitle drafts.

Excel or Google Sheets: Used for content data statistics.

Midjourney Or it could be a dream: used for concept art and cover design reference, but the official commercial image must be confirmed by the designer.

Grammarly, or Chinese proofreading tool, is used for checking for typos, grammatical errors, and punctuation.

Suggested process:

Topic selection stage: The content manager determines the topic, target audience, and core viewpoints.

Data stage: For primary editors AI When compiling data summaries, source links must be retained.

Outlining stage: Senior editors determine the article structure and main viewpoints.

First draft stage:AI The tool helps generate paragraph drafts, which the author then reviews and rewrites.

Review phase: Use the fact-check checklist, brand tone checklist, and sensitive expression checklist.

Reuse phase: Break down long articles into short video scripts, short articles for social media, and sales scripts.

Post-mortem phase: Operations and data analysis collect reading, conversion, and completion rate data weekly.

Job division suggestions:

Content manager: Determines topic priorities, reviews core viewpoints, and approves publication.

Senior Editor: Responsible for judging articles, outlining, and finalizing revisions.

Junior Editor: Responsible for data organization, draft generation, and source attribution.

Short video director: Break down long articles into scripts, narration scripts, and storyboards.

Designer: Responsible for visual style, cover image, and infographics.

Operations: Responsible for publishing schedules, distribution channels, and managing comment feedback.

Data Analysis: Responsible for tracking results, monthly reports, and topic selection suggestions.

Risk control requirements:

All facts, data, and citations must be manually verified.

AI The generated content cannot be published directly.

When dealing with customer names, contract information, or sales data, public information must not be entered. AI tool.

Product capability descriptions must be confirmed by the product team or pre-sales department.

Content from sensitive industries such as healthcare, finance, and law requires additional review.

all AI When generating images, copyright and commercial risks must be checked.

The final viewpoint of an article must be confirmed by the author; it cannot be merely adopted from others. AI in conclusion.

Budget:

The pilot program has a budget of 90,000 yuan and a duration of 3 months.

Budget Breakdown:

AI Tool account: 30,000 yuan.

Internal training andPrompt wordsTemplate construction: 20,000 yuan.

Knowledge base organization and process configuration: 15,000 yuan.

Design and video auxiliary tools: 15,000 yuan.

Reserved funds: 10,000 yuan.

Schedule:

Week 1: Confirm the scope of the pilot program, select the tools, and establish account permissions.

Week 2: Organize frequently used itemsPrompt wordsTemplates and audit checklists.

Weeks 3 to 4: Select two content types, WeChat official account articles and customer case studies, for a small-scale pilot program.

Weeks 5 through 8: Expand to short video scripts and sales materials, and begin weekly debriefings.

Weeks 9 to 10: Adjust processes based on data and supplement risk control rules.

Weeks 11-12: Develop a formal SOP, output a pilot report and budget recommendations for the next phase.

Pilot scope:

The first phase only covers 4 types of content:

In-depth articles on WeChat official accounts.

Client case studies.

Short video script.

Sales support materials.

Not included for the time being:

Article signed by senior executives.

Press release for a major brand.

Content involving undisclosed customer data.

Industry reports with high legal risks.

Measurement metrics:

Monthly content output.

Average production cycle for a single piece of content.

Data processing takes time.

First draft rework rate.

Title modification rounds.

Read completion rate within 7 days of publication.

The sales team rates the availability of the materials.

AI Number of issues discovered during content review.

Expected results:

After a 3-month pilot program, the content team should develop a reusable set of procedures. AI Work process.AI The main responsibilities include data organization, drafting structures, selecting alternative titles, reusing content, and initial data screening. Core viewpoints, fact-checking, client feedback, and final publication decisions are still handled manually.

The task requirements are met; it's clear that the understanding of the project content is thorough, and the aesthetics are also acceptable. It's already usable; with a little manual fine-tuning, it can be submitted directly for reporting.

After testing with GLM-5.2, I feel that the previous rankings might actually be quite accurate. I think the results for the tasks are all impressive, representing the top level of domestic models. They not only run and function well, but they are also very user-friendly.

In the past, when I was working, I had to work on things like organizing data, creating initial versions of code, building pages, structuring PPTs, and test cases, all of which had to be refined bit by bit. Now, I can let the model run a version first, and I can focus my energy back on judgment and selection.

My assessment of GLM-5.2 is: it has a high ceiling and stable performance, making it worthwhile to seriously try it out in my daily workflow. As for whether it can be retained long-term, we'll have to see how the model performs in terms of detail, stability, and cost during high-frequency use in the coming weeks.

When connecting to APIs in the future, it might be impossible to distinguish whether it's GLM-5.2 or... Claude It's over.

Original link:GLM-5.2 Real-world testing shows that Chinese models have truly reached the top tier.