Real, verified AI agent use cases
142 real examples of what people have gotten AI agents to do (including 18 failures and partial results), each with the prompt they used and a link to the original source. Copy a prompt and try it, or search for what you're trying to get done.
Recreated prompt = reconstructed by us · How we verify
Not sure where to start?
142 of 142 cases, newest first
- Log new board game purchases and add them to BoardGameGeek with Claude in ChromeEvidence: Screenshots or logs
From one instruction, the agent found the ordered games, created notes, downloaded cover images, added metadata and added the games to the poster's BoardGameGeek account using Claude in Chrome.
Claude in Chrome · Personal · Source date:
- Find flights with Manus (16-hour run, poor pick)Recreated promptEvidence: Described by posterFailure
Asked to find flights, Manus ran for 16 hours and then picked what the poster called probably the most expensive option, with a 9-hour layover.
Manus · Travel · Source date:
- Write a drama-style ad script modeled on a winning ad with ManusRecreated promptEvidence: Described by posterBeginner-friendly
Manus wrote a drama-format script for one of the brand's top historical winning ad concepts, modeled on a winning drama ad from a beauty brand; the poster says it hit 5 ROAS right away.
Poster says:5 ROAS off the bat
Manus · Marketing · Source date:
- Write the next gym training cycle from workout logs with Claude CodeRecreated promptEvidence: Screenshots or logs
Claude Code read four weeks of the poster's workout logs, found where lifts stalled, wrote the next training cycle with a reason for each starting weight, loaded the routines into the tracking app, and noted where the app would suggest the wrong weight.
Claude Code · Personal · Source date:
- Clip every digital coupon on a grocery account with Claude CoworkRecreated promptEvidence: Described by poster
Claude Cowork logged into the poster's Jewel-Osco account and clipped every available coupon.
Claude Cowork · Shopping · Source date:
After Copilot suggested a subflow that didn't exist in Power Automate Desktop, the poster asked Claude Cowork, which took over and built the flow on their machine.
Claude Cowork · Ops · Source date:
- Rename a pile of train tickets by date and destination with Claude CoworkRecreated promptEvidence: Described by poster
Claude Cowork read a pile of train tickets and renamed each file by its correct date and destination.
Claude Cowork · Personal · Source date:
- Submit a hosting provider bug report from log warnings with Claude in ChromeRecreated promptEvidence: Screenshots or logs
Using the poster's logged-in browser, Claude in Chrome submitted a bug report to DigitalOcean after finding warnings in the logs.
Claude in Chrome · Ops · Source date:
- Plan a six-week Southeast Asia trip with Manus (hallucinated flights)Recreated promptEvidence: Described by posterFailure
The poster used Manus to plan a 6-week Southeast Asia trip, and it hallucinated flights that didn't exist despite having browser access.
Manus · Travel · Source date:
- Split, rename and sort about 200 scanned invoices with Claude CoworkRecreated promptEvidence: Described by poster
After the poster scanned almost 200 invoices, Claude Cowork split them out, renamed them, sorted them into folders and created a summary Excel file.
Claude Cowork · Ops · Source date:
- Find car brands without a web app and email them a pitch with ManusRecreated promptEvidence: Screenshots or logs
Manus searched for car brands without a web app, wrote an email and sent it to them; some replied and the poster built one an app.
Manus · Marketing · Source date:
- Turn an inbox into a daily to-do list with Claude Cowork (abandoned)Recreated promptEvidence: Described by posterFailure
Claude Cowork summarized the poster's email into a to-do list, but the list grew long and unmanageable, so the poster stopped reading it and turned it off as a waste of tokens.
Claude Cowork · Personal · Source date:
Muse booked every restaurant for the poster's Japan trip and corresponded in Japanese with a Hakone hotel to secure a dinner slot not offered on its website.
Muse · Travel · Source date:
- Shop for cheaper home insurance quotes with Claude Code and ChromeRecreated promptEvidence: Described by poster
Using the poster's coverage details and access to Chrome, Gmail and the existing policy, Claude Code sought quotes from about a dozen insurers and found two strong candidates; the poster says they were approved by AAA at $3,100/year, a $4,600/year saving.
Poster says:$4,600/year saved
Claude Code, Claude in Chrome · Personal · Source date:
- Find, negotiate and review a small office lease with CodexRecreated promptEvidence: Described by poster
Codex found a 74 sq ft commercial lease near the poster's kids' school, negotiated the terms and reviewed the contract; the poster found it cheaper than a coworking desk.
Codex · Ops · Source date:
- Book a whole weekend trip in 30 minutes with ChatGPT WorkRecreated promptEvidence: Described by poster
ChatGPT read the poster's chat with their girlfriend, proposed itinerary options, staged an Airbnb, rental car, fishing gear and dinner reservations through browser use, then made a shareable trip page after the poster booked.
ChatGPT Work · Travel · Source date:
- Manage Google Ads with Claude Code connected to analytics and the websiteRecreated promptEvidence: Described by poster
With Google Ads, Google Analytics and the Webflow site connected to a Claude Code project, the poster adds negative keywords quickly, tracks ad performance, and found that mobile wasn't converting because the page showed nothing relevant above the fold.
Claude Code · Marketing · Source date:
- Screen and clear an email inbox in one pass with Claude CodeRecreated promptEvidence: Described by poster
The poster runs all email in Claude Code: in one pass it screened 27 emails, archived and unsubscribed, and sent sub-agents after the three that needed real thought.
Claude Code · Personal · Source date:
- Prepare an Indian income tax return (ITR-3) with Claude CodeRecreated promptEvidence: Described by poster
Given tax documents and a year of bank statements, Claude Code reconciled the figures against the poster's AIS (annual information statement), caught 2 mistakes that could have triggered notices, and computed the tax; the poster then filed.
Poster says:2 mistakes caught
Claude Code · Ops · Source date:
- Organize a Spanish visa renewal and fill the forms with Claude and Claude in ChromeRecreated promptEvidence: Described by poster
For a Spanish visa renewal, the poster set up a bureaucracy folder with Claude, fed in the family's information, had Claude verify documents, then used Claude in Chrome to fill the online forms faster and more accurately than checking by hand.
Claude, Claude in Chrome · Personal · Source date:
- Build a work-trip itinerary from booking emails with ClaudeRecreated promptEvidence: Described by poster
The day before a work trip, Claude read the poster's booking emails and organized a polished itinerary that the poster says would have taken a few hours by hand.
Claude · Travel · Source date:
- Qualify scraped leads against your ideal customer before cold email with Claude CodeRecreated promptEvidence: Described by poster
Claude Code reads each scraped company's website and checks it against the poster's ideal customer profile, dropping about 30% of leads before outreach; the poster says reply rates roughly doubled and they booked 100+ meetings in 3 months.
Poster says:100+ meetings booked in 3 months
Claude Code · Marketing · Source date:
- Write and publish a short podcast episode with Claude Code (needed edits)Recreated promptEvidence: Described by posterPartial
Claude Code wrote a short podcast script that the poster voiced with ElevenLabs and published via Anchor; the narration was clear but the script missed nuance and needed manual tweaks, so the poster rated it useful for drafts only.
Claude Code · Content · Source date:
- Transcribe and organize 411 family voice memos with Claude CodeRecreated promptEvidence: Screenshots or logs
Claude Code transcribed and clustered 411 voice memos left by the poster's late grandmother and built a small site organized by mood and date, which the family uses more than the photo album.
Poster says:411 voice memos transcribed
Claude Code · Personal · Source date:
- Reconcile crypto trades across chains and post them to Xero with Claude CodeRecreated promptEvidence: Described by poster
Claude Code tracked on-chain swaps across 10+ assets, produced a monthly gain/loss breakdown with reconciled exchange rates, then built a Xero connection and posted journal entries.
Claude Code · Ops · Source date:
- Turn a year of bank statements into a business expense spreadsheet with Claude CoworkRecreated promptEvidence: Described by poster
Claude Cowork extracted and categorized business expenses from a year of statements across several accounts into a well-organized spreadsheet; after some back-and-forth the poster fixed a handful of items by hand.
Claude Cowork · Data · Source date:
- Extend a family tree two generations with Claude Code and browser agentsRecreated promptEvidence: Shown on video
Working from the family's archive scans, Claude Code with memory and the Claude browser plugin cross-checked church books, ship manifests and naturalization records, built a 300-page family wiki, and found two more generations in a day; the family is now checking it for errors.
Poster says:300-page family wiki
Claude Code, Claude in Chrome, Perplexity Comet · Research · Source date:
- Match a tax-prep checklist against a receipts folder with Claude CoworkRecreated promptEvidence: Described by poster
The poster says matching a spreadsheet of items for their accountant against a folder of receipts usually took 4–6 hours, and that with Claude Cowork it took about an hour.
Poster says:4–6 hour job done in about an hour
Claude Cowork · Ops · Source date:
- Book a doctor's appointment with agent mode (blocked)Recreated promptEvidence: Described by posterFailure
The poster tried to set up a doctor's appointment with agent mode and got nowhere; they say ChatGPT and Manus agents are held back by blocking and captchas.
ChatGPT agent · Personal · Source date:
- Draft a firm landlord notice about a repeat water leak with Claude CodeRecreated promptEvidence: Described by poster
Instead of paying a lawyer for a strongly worded notice, the poster gave Claude Code the lease documents and context about a second water leak; it drafted an email that got them the concessions they wanted.
Claude Code · Personal · Source date:
- Make years of calls, podcasts and posts searchable with Claude CodeRecreated promptEvidence: Described by poster
Claude Code organized years of calls, podcasts, tweets and presentations into a searchable database connected to AI, with a daily job that adds new content; asked for a masterclass outline, it drew on 224 chunks from 78 sources.
Claude Code · Content · Source date:
- Migrate two websites from WordPress to Wagtail with Claude CodeRecreated promptEvidence: Described by poster
Over a weekend Claude Code migrated two sites from WordPress to Wagtail (PHP to Python), including server setup, data transfer, testing and fixes, and redesigned a homepage; the poster reports nothing broke and traffic and tools stayed in place.
Claude Code · Coding · Source date:
- Prepare and file overdue German VAT returns with Claude Cowork and Claude in ChromeRecreated promptEvidence: Shown on video
From a folder of invoice and receipt PDFs, Claude Cowork prepared quarterly accounts; after the poster logged in to the Elster tax portal, Claude in Chrome found, filled and submitted the Q3 2025 VAT forms.
Claude Cowork, Claude in Chrome · Ops · Source date:
- Find smarter buying patterns in office supply order history with Claude CodeRecreated promptEvidence: Described by poster
Alex Hillman had Claude Code go through Indy Hall's paper products order history to suggest buying patterns that avoid costly last-minute orders.
Claude Code · Data · Source date:
- Analyze seven years of credit card statements with Claude CoworkRecreated promptEvidence: Screenshots or logs
From seven years of monthly credit card statement PDFs in zip files, Claude Cowork returned an Excel file with 5,726 extracted, categorized transactions plus analysis; issues were vendor transaction IDs and some miscategorized vendors it wasn't told how to handle.
Poster says:5,726 transactions categorized
Claude Cowork · Data · Source date:
- Turn six months of receipts into a categorized spreadsheet with Claude CoworkRecreated promptEvidence: Shown on video
Pointed at a folder, Claude Cowork organized 6 months of receipts into a categorized spreadsheet with monthly breakdowns.
Claude Cowork · Ops · Source date:
- Plan a family Disneyland trip with Claude Code and SkillsRecreated promptEvidence: Screenshots or logs
The poster planned a family Disneyland trip using Claude Code as the agent, Claude Skills for research, planning, tracking and journaling, and a simple companion web app for itineraries and maps.
Claude Code · Travel · Source date:
- Do year-end bookkeeping and Xero reconciliation with Claude CodeRecreated promptEvidence: Screenshots or logs
Claude Code categorized a folder of receipts, cross-referenced exported bank CSVs against Xero to find duplicates and missing transactions, used the browser extension to report on reconciliation items, and produced a cost-cutting report; the poster says this replaced roughly $2–3K of accounting-firm work.
Poster says:~2–3K accounting-firm fee avoided
Claude Code, Claude in Chrome · Ops · Source date:
- Find hard-to-locate company tax IDs and addresses with GPT-5 ProEvidence: Described by poster
Working for over 45 minutes, GPT-5 Pro found correct FEINs for 4 of 6 subsidiaries on the Florida Division of Corporations site and one address via a FINRA BrokerCheck report; the poster double-checked the FEINs.
GPT-5 Pro · Research · Source date:
- Apply to LinkedIn jobs from a resume with Perplexity CometRecreated promptEvidence: Screenshots or logs
Comparing Comet and ChatGPT Atlas for LinkedIn job applications, the poster found Comet far better; with a detailed prompt and their resume, it applied to 20+ jobs and sent notes to HR.
Poster says:20+ job applications
Perplexity Comet · Personal · Source date:
- Spot when you avoid conflict in meeting recordings with Claude CodeRecreated promptEvidence: Article or interview
Dan Shipper puts all his meeting recordings in a folder and asks Claude Code to find every time he subtly avoided conflict.
Claude Code · Personal · Source date:
- Diagnose why your Mac is running slow with Claude CodeRecreated promptEvidence: Article or interview
Anthony Roux uses Claude Code to check load, memory pressure, disk space, stuck processes and swap, dig into causes like caches, Docker and Time Machine snapshots, explain the results and suggest fixes with risk notes.
Claude Code · Personal · Source date:
- Port a backend to Convex in four hours with Factory's DroidRecreated promptEvidence: Described by poster
The poster reports that Factory's Droid ported their tRPC/Drizzle custom backend to Convex in about 4 hours, with some bugs left to fix.
Factory Droid · Coding · Source date:
- Migrate a 50-screen app from Firebase to Supabase with Claude CodeRecreated promptEvidence: Described by poster
Going feature by feature, the poster used Claude Code to generate database schemas, write queries and fix daily errors; they report migrating about 85% of a 45–50 screen Flutter app from Firebase to Supabase in 10–20 days.
Poster says:~85% migrated in 10–20 days
Claude Code · Coding · Source date:
- Notify Germany's TV licence (GEZ) office of a house move with ChatGPT agentRecreated promptEvidence: Described by poster
After a few problems, ChatGPT agent found the GEZ (public broadcasting fee) contact form, wrote a message and entered the new address; the poster only had to solve one captcha.
ChatGPT agent · Personal · Source date:
- Find a restaurant-week steak deal and book a table for 2 with ChatGPT agentEvidence: Described by poster
Asked to find the best Dine LA steak dinner deal and book for 2 the next day, ChatGPT agent booked Dear John's for 8:15, though it misread the menu by treating sides as a separate course.
ChatGPT agent · Personal · Source date:
- Complete a 31-step web form: ChatGPT agent vs Perplexity CometRecreated promptEvidence: Described by posterPartial
On a 31-step form with buttons, dropdowns and fields, Comet failed after 3–5 steps, while ChatGPT agent completed all 31 and corrected its own mistakes by checking against the steps.
ChatGPT agent, Perplexity Comet · Ops · Source date:
- Turn one video script into a newsletter, titles, thread and description with Claude Code subagentsRecreated promptEvidence: Screenshots or logs
From one long-form video script, an orchestrator in Claude Code calls subagents that write a newsletter, generate titles, create thumbnails, write a thread and draft the YouTube description.
Claude Code · Content · Source date:
- Pay a UK car tax reminder with ChatGPT agent (handoff for payment)Recreated promptEvidence: Described by poster
From a photo of the posted reminder and a chosen payment option, ChatGPT agent found the government website and filled out the forms; the poster took over to enter card details.
ChatGPT agent · Personal · Source date:
- Prep an Instacart order from history and sales with ChatGPT agent (odd picks)Recreated promptEvidence: Described by poster
Given access to Instacart and told to prep an order from the poster's history and current sales, ChatGPT agent picked items with reasoning the poster found oddly generic ("cereal, which is on sale and a usual part of groceries").
ChatGPT agent · Shopping · Source date:
- Plan a points-funded family holiday and book dinner with ChatGPT agent (partial)Recreated promptEvidence: Described by posterPartial
For a points-funded family holiday plan, ChatGPT agent built a decent deck but spent 15 minutes formatting slides; it couldn't book the romantic dinner.
ChatGPT agent · Travel · Source date:
- Apply to a job with ChatGPT agent (small-button trouble)Recreated promptEvidence: Shown on videoPartial
Testing ChatGPT agent on a real job application (to OpenAI), the poster found it much better than the earlier Operator but it struggled to click small buttons; they rated Manus as slightly ahead at times.
ChatGPT agent · Personal · Source date:
- Check a hotel price with ChatGPT agent (wrong town)Recreated promptEvidence: Described by posterFailure
Asked to check a hotel's price with its town clearly stated, ChatGPT agent looked up a same-named hotel in another town, and the poster notes this wouldn't have been obvious from the reply.
ChatGPT agent · Travel · Source date:
- Order groceries from H-E-B with ChatGPT agent (access denied)Recreated promptEvidence: Screenshots or logsFailure
The poster tried to order groceries from H-E-B in Austin with ChatGPT agent, but it failed because access was denied.
ChatGPT agent · Shopping · Source date:
- Turn 200 photos of household items into a moving-sale list with Claude CodeRecreated promptEvidence: Described by poster
Claude Code processed over 200 photos of the poster's household items and created a moving-sale list with descriptions and prices; the poster says it saved hours compared with writing Facebook Marketplace listings by hand.
Claude Code · Shopping · Source date:
- Compare prices for 60–70 garden plants across 5 sites with OpenAI OperatorRecreated promptEvidence: Screenshots or logs
Overnight, Operator cross-referenced a list of 60–70 plants across 5 sites into a price comparison table; the poster called it buggy but fun.
OpenAI Operator · Shopping · Source date:
- Batch-enter design prompts into an image generator: Operator vs Project MarinerRecreated promptEvidence: Screenshots or logsPartial
For an Etsy design shop, Operator entered a ChatGPT-made list of prompts into Ideogram to generate draft designs after the user logged in; Project Mariner couldn't get past Ideogram's captcha even with the user in control.
OpenAI Operator, Project Mariner · Content · Source date:
- Find links to content with Project Mariner (no links returned)Recreated promptEvidence: Described by posterFailure
Asked to find links to content, Project Mariner found the content but wouldn't return the links, which the poster said made the result not very helpful.
Project Mariner · Research · Source date:
Talking to Codex in voice mode while cooking dinner, the author got a Django newsletters archive with four importers and a pull request, then finished the rest with typed prompts.
OpenAI Codex (ChatGPT desktop, voice mode) · Coding · Source date:
- Submit a product to ~300 directories and launch sites with Claude in ChromeRecreated promptEvidence: Described by poster
In one day, Claude in Chrome filled directory and launch-platform forms 5 tabs at a time; 134 of about 300 submissions went through, with the human handling logins and captchas.
Poster says:134 submissions went through
Claude in Chrome · Marketing · Source date:
- Fix a Product Hunt X handle across launches and comments with a browser agentRecreated promptEvidence: Screenshots or logs
A browser agent corrected the poster's X handle from johnrushx to johnrush across all Product Hunt launches and hundreds of comments.
browser agent (Claude in Chrome) · Marketing · Source date:
- Migrate 180 microservices to a new framework version with Cursor agentsRecreated promptEvidence: Company case study
Brex reports that its self-correcting, self-hosted Cursor agents let one engineer migrate 180 microservices from Micronaut 3 to 4 in a month, which it calls a 32x per-capita speedup.
According to Brex:180 microservices migrated in one month
Cursor (self-hosted agents) · Coding · Source date:
- Mine Gmail for on-record PR quotes with Claude CoworkEvidence: Article or interview
Connected read-only to Gmail, Claude Cowork found 12 relevant PR pitches in about a minute, then narrowed to eight on-record quotes with publish permission that the journalist verified before use.
Claude Cowork · Research · Source date:
- Automate defect remediation across C/C++ codebases with Codex CLIRecreated promptEvidence: Company case study
According to OpenAI's case study, Cisco's Codex CLI defect-repair workflow raised defect resolution throughput 10–15x, and Codex build optimization saved more than 1,500 engineering hours a month.
According to OpenAI:10–15x defect resolution throughput
OpenAI Codex CLI · Coding · Source date:
- Let a whole team self-serve data via an agent-written schema guideRecreated promptEvidence: Described by poster
Claude Code wrote a markdown guide to the company's data model and business terms from SQLMesh context and past queries; Hex's agent now answers the team's data requests, and the analyst hire was dropped.
Claude Code, Hex agent · Data · Source date:
- Audit 90 days of Meta ads from a CSV export with Claude CodeEvidence: Described by poster
Claude Code flagged 14 money-losing creatives, found that 7 of the top 10 ads opened with a question, and spotted 3 ad sets cannibalizing each other.
Poster says:14 money-losing creatives flagged
Claude Code · Marketing · Source date:
A weekend-built Claude Code pipeline scrapes leads, scores them against the ICP, writes spintax copy and launches the Instantly campaign with one command; the poster says it saves about 2 hours each morning.
Poster says:2 hours back every morning
Claude Code · Marketing · Source date:
- Merge provider price CSVs into one comparison table with Claude CodeRecreated promptEvidence: Described by poster
Claude Code read several local CSVs of model prices from different providers and produced one organized comparison table grouped by provider.
Claude Code · Data · Source date:
- Turn podcast highlights into timestamped clips with Claude Cowork + Claude in Chrome + CodexRecreated promptEvidence: Described by poster
Claude Cowork expanded the poster's top 10 podcast takeaways, then used Claude in Chrome to find the transcript and timestamps for each clip, and Codex downloaded them.
Claude Cowork, Claude in Chrome, OpenAI Codex · Content · Source date:
- Enter retail-arbitrage receipts into a spreadsheet with Claude CoworkRecreated promptEvidence: Described by poster
Claude Cowork weekly processes a folder of RA receipts into a spreadsheet with buy cost, store and item; the poster still double-checks but says admin time is way down.
Claude Cowork · Ops · Source date:
- Automate a daily analytics summary to Slack with Claude CodeRecreated promptEvidence: Described by poster
Claude Code set up a GitHub Actions cron that pulls PostHog data at 8am, summarizes it in Python and posts to Slack; the first report arrived the next morning.
Claude Code · Data · Source date:
- Draft government-solicitation proposals in 25 minutes with n8n AI workflowsRecreated promptEvidence: Company case study
According to n8n's case study, Field Aerospace's self-hosted n8n AI workflows bring a proposal draft that used to take about two weeks of work from three or four people to an 80% starting point in roughly 25 minutes, and replaced two legacy tools, eliminating approximately $30,000 in annual software costs (the case study's headline gives $22,000 a year).
According to n8n:Proposal time cut from 2 weeks to 25 minutes
n8n (AI-powered workflows) · Ops · Source date:
- Build a weekly business KPI dashboard workflow with Claude Code + n8nRecreated promptEvidence: Article or interview
Claude Code built a complex n8n workflow for a client's weekly KPI dashboard in about 15 minutes; the workflow then runs unattended on n8n.
Claude Code, n8n · Ops · Source date:
- Break writer's block on a book with Claude Code interviewsRecreated promptEvidence: Described by poster
Claude Code reviewed a book draft, interviewed the author to fill gaps, and adapted his dictated answers into the manuscript in his own voice.
Claude Code · Content · Source date:
Claude Code installed yt-dlp, pulled and cleaned the captions of an 84-minute Gurley/Gerstner/Nadella talk, wrote focused debriefs, then packaged it as a reusable skill.
Claude Code · Content · Source date:
- Turn X bookmarks and likes into a searchable knowledge baseRecreated promptEvidence: Described by poster
The poster scraped all their X bookmarks and likes with Claude in Chrome and other tools, cleaned the linked pages with Firecrawl, and built a RAG chatbot over everything they had saved.
Claude in Chrome, Vercel Agent Browser · Data · Source date:
- Analyze every entry of a huge JSONL file with Claude Code subagentsRecreated promptEvidence: Described by poster
Claude Code previewed the data, trimmed fields, defined a schema and ran hundreds of Haiku subagents over ~45 minutes so every entry in a large JSONL file was analyzed.
Claude Code · Data · Source date:
- File Thailand Digital Arrival Cards for five travelers with ChatGPT agentRecreated promptEvidence: Described by poster
ChatGPT agent filled Thailand Digital Arrival Cards for 5 people (about an hour / 16 credits); the poster verified the preview and swapped in real passport numbers before submitting.
ChatGPT agent · Travel · Source date:
- Hunt last-minute holiday flights with multiple agents (failure case)Recreated promptEvidence: Described by posterFailure
Claude Opus 4.5, Gemini 3 Pro and ChatGPT agent mode only matched a manual Google Flights search and were too slow as cheap fares vanished; the humans beat them by about 15%.
ChatGPT agent, Claude Opus 4.5 (Antigravity), Gemini 3 Pro · Travel · Source date:
- Port an HTML5 parser from Python to JavaScript with Codex CLIEvidence: Screenshots or logs
Codex CLI (GPT-5.2) ran for hours largely unattended and produced about 9,000 lines of dependency-free JavaScript passing 9,200 html5lib tests, in about 4.5 hours start to finish.
Poster says:Passes 9,200 html5lib tests
OpenAI Codex CLI · Coding · Source date:
- Auto-coach sales reps from call recordings with Zapier AgentsRecreated promptEvidence: Company case study
According to Zapier's customer story, Healthie's call-summary and coaching agents save each of about 20 reps and CSMs 2–3 hours a week, more than 60 hours a week across the team.
According to Zapier:60+ hours a week saved
Zapier Agents · Ops · Source date:
- Compare Claude Code, Codex and Gemini CLI on a 97k-message chat datasetRecreated promptEvidence: Described by poster
The poster gave the same 97k-message WeChat dataset to Claude Code, Codex and Gemini CLI as analysts and published results with an evaluation table.
Claude Code, OpenAI Codex CLI, Gemini CLI · Data · Source date:
- Run async code-research benchmarks with a coding agentEvidence: Screenshots or logs
An async coding agent cloned several Python Markdown libraries, benchmarked them against cmarkgfm, and wrote a report with charts, with almost no human time spent.
Claude Code for web, OpenAI Codex · Research · Source date:
- Upgrade a plugin to a new framework alpha with Codex CLIEvidence: Screenshots or logs
A one-line `codex exec` loop read the Datasette 1.0a20 upgrade docs, fixed the failing tests, and got the datasette-checkbox plugin passing again.
OpenAI Codex CLI · Coding · Source date:
An n8n agent researches top AI news via Brave Search, generates content from templates, and emails a weekly idea pack.
n8n agent · Content · Source date:
On the same Airbnb search, ChatGPT Atlas agent mode and Comet each found properties in about 5 minutes, Claude's extension took 10 minutes to find the first, and MiniMax Agent ran over an hour without giving links.
ChatGPT Atlas agent mode, Perplexity Comet, Claude in Chrome, MiniMax Agent · Travel · Source date:
- Build a terminal-to-HTML sharing tool with Claude Code for webEvidence: Shown on video
Claude Code for web built a terminal-to-html tool that turns pasted terminal output into a shareable gist-preview link, recorded start to finish in an 11-minute video.
Claude Code for web · Coding · Source date:
- Build a B2B SaaS press list of 40–50 journalists with ManusEvidence: Article or interviewBeginner-friendly
Manus produced a Notion-ready press list of journalists covering mid-market SaaS/enterprise software with contacts, recent articles and pitch angles; the author says work that used to take ~6 hours took about 25 minutes, with some contact verification still needed.
Reported by The AI Enterprise:6 hours of work done in 25 minutes
Manus · Marketing · Source date:
- Generate 60+ AI-consulting sales prospects with ManusEvidence: Article or interviewBeginner-friendly
A Manus prompt for mid-market AI-implementation prospects returned about 73 qualified companies with decision-maker contacts and outreach notes in a Notion/spreadsheet export.
Reported by The AI Enterprise:73 prospects found
Manus · Marketing · Source date:
- Plan St. Simons Island restaurants with ChatGPT agentEvidence: Article or interviewBeginner-friendly
ChatGPT Agent Mode produced a dining guide for a 4-day St. Simons Island trip organized by meal and price, and the author ate at 4 restaurants from the list and liked them all.
ChatGPT agent · Travel · Source date:
- Find Southeast AI events for meetup promotion with ManusEvidence: Article or interviewBeginner-friendly
Manus found 34 upcoming AI/tech events in North Carolina and the Southeast suitable for promoting a meetup, exported to Notion with contacts and opportunity type, cutting an afternoon of work to about 20 minutes.
Reported by The AI Enterprise:34 relevant events found
Manus · Marketing · Source date:
- Get DeepSeek-OCR running on an NVIDIA Spark with Claude CodeEvidence: Screenshots or logs
With four prompts and about 40 minutes, Claude Code figured out the right ARM CUDA PyTorch wheels, got DeepSeek-OCR running on NVIDIA Spark hardware, and OCR'd a test image.
Claude Code · Coding · Source date:
- Sort and rename messy invoices for taxes with Claude CodeRecreated promptEvidence: Article or interview
Claude Code reads each file in a messy invoice folder, renames it to 'YYYY-MM-DD Vendor - Invoice - ProductOrService.pdf' and files it in the right folder.
Claude Code · Ops · Source date:
- Turn stroller-walk voice notes into a published article with Claude CodeRecreated promptEvidence: Article or interview
A non-developer fed rambling voice notes to Claude Code, which organized them into research themes, wrote a full article in her voice and made LinkedIn versions.
Claude Code · Content · Source date:
- Find working coupon codes at checkout with ChatGPT agentRecreated promptEvidence: Screenshots or logs
ChatGPT agent kept trying coupon codes until it found ones that worked; the poster says it saved them $79.
Poster says:$79 saved
ChatGPT agent · Shopping · Source date:
- Find local takeout deals under a budget with ChatGPT agentEvidence: Article or interviewBeginner-friendly
The Tom's Guide author uses ChatGPT agent to scour delivery apps and local restaurants for family takeout deals under a set budget.
ChatGPT agent · Shopping · Source date:
- Track and cancel unused subscriptions with ChatGPT agentEvidence: Article or interview
ChatGPT agent scans for recurring charges and reminds the author to cancel before renewal, helping cancel unused streaming services.
ChatGPT agent · Personal · Source date:
- Write and ship a complete Datasette plugin with Claude CodeEvidence: Screenshots or logs
Claude Code (Sonnet 4.5) wrote every line of code and every test for the datasette-os-info plugin with minimal prompting, and the author shipped it.
Claude Code · Coding · Source date:
- Reproduce an economics paper's findings from replication data with ClaudeEvidence: Article or interview
Given only the paper and its replication archive, Claude Sonnet 4.5 converted the STATA code to Python and reproduced the findings; the author spot-checked the results and GPT-5 Pro confirmed them.
Claude (Sonnet 4.5) · Data · Source date:
A scheduled ChatGPT agent checks every day whether a product is in stock, reports at 8am, and even flags status changes like 'out of stock' becoming 'coming soon'.
ChatGPT agent · Shopping · Source date:
- Weekly X analytics review and growth coaching with ChatGPT agentRecreated promptEvidence: Screenshots or logs
Every Monday ChatGPT agent pulls the poster's X analytics, studies which posts worked, compares profile visits with follower growth, and suggests bio and pinned-post changes.
ChatGPT agent · Marketing · Source date:
- Compare top social media tools in a spreadsheet with ChatGPT agent and ClaudeRecreated promptEvidence: Screenshots or logsBeginner-friendly
Allie K. Miller asked both ChatGPT Agent Mode and Claude's Excel feature to compare top social media tools, and Claude finished in 3 minutes vs 8 with denser decision support, though some prices mixed annual vs monthly.
ChatGPT agent, Claude (Excel) · Research · Source date:
- Categorize 419 ledger transactions into an Excel template with ChatGPT agentRecreated promptEvidence: Screenshots or logs
Given an Excel ledger template and 419 transactions, ChatGPT agent asked for help on ambiguous merchants, then keyed in all ~400 rows and took 4 minutes.
Poster says:Done in 4 minutes
ChatGPT agent · Ops · Source date:
- Allocate a 30-employee payroll journal across departments with ChatGPT agentRecreated promptEvidence: Screenshots or logs
ChatGPT agent mapped a payroll CSV into a 30-employee allocation spreadsheet across 6 departments, making 180 entries and summing employer-paid columns in under 4 minutes.
Poster says:180 entries in under 4 minutes
ChatGPT agent · Ops · Source date:
- Find the best flight options on Skyscanner with ChatGPT agentEvidence: Screenshots or logs
ChatGPT agent searched Skyscanner across flexible dates and returned the flight the author would actually book in 3 of 3 tests, though airline sites and 3D Secure blocked it from completing the booking.
ChatGPT agent · Personal · Source date:
- Make a UK tax research agent from a folder of HMRC manuals with Claude CodeRecreated promptEvidence: Screenshots or logs
Using scraped HMRC manuals and legislation plus a CLAUDE.md that launches parallel subagents, Claude Code answered an ATT exam question in about 5 minutes, earning roughly 2.5 of 3 marks by the author's grading.
Claude Code · Research · Source date:
- Search for cheap flights with ChatGPT agent (failure case)Recreated promptEvidence: Described by posterFailure
ChatGPT agent suggested flights costing $1,400 when the poster found an $850 option on their own.
ChatGPT agent · Travel · Source date:
- Find an available domain name with ChatGPT agentRecreated promptEvidence: Described by posterBeginner-friendly
The poster found ChatGPT agent useful for finding an available domain name that met a set of constraints.
ChatGPT agent · Personal · Source date:
A solo creator had ChatGPT agent log into her Google Ads account and fix a conversion-tracking issue.
ChatGPT agent · Marketing · Source date:
- Book a business-lunch restaurant reservation with ChatGPT agentEvidence: Article or interview
From a terse prompt, ChatGPT agent found a restaurant, filled in the form (including a shellfish allergy), and got a confirmed reservation in 11 minutes versus about 2 manually.
ChatGPT agent · Personal · Source date:
Comet Assistant created a Steam account with the email the poster gave it, and solved a captcha the poster had failed.
Perplexity Comet Assistant · Personal · Source date:
ChatGPT agent logged into Gym Meals Direct, chose meals and sizes, picked Monday delivery and applied a discount.
ChatGPT agent · Shopping · Source date:
- Book a restaurant reservation with ChatGPT agent (failure case)Recreated promptEvidence: Described by posterFailure
ChatGPT agent took about 5 minutes just to reach the reservation page, then handed back; the poster called it a waste of time.
ChatGPT agent · Personal · Source date:
- Find the cheapest one-way flight across two departure cities with ChatGPT agentEvidence: Screenshots or logsBeginner-friendly
From a vague brief, ChatGPT agent searched flights from Ljubljana or Zagreb to Bordeaux over a two-week window and found a €141.48 fare in 12 minutes.
ChatGPT agent · Travel · Source date:
- Draft and complete an eBay Germany listing from photos with ChatGPT agentRecreated promptEvidence: Shown on video
From a few photos, ChatGPT agent identified an Apple Watch Series 7, noted display scratches, drafted a German title and description, then finished the eBay DE listing draft after login in about 15 minutes.
ChatGPT agent · Shopping · Source date:
- Weekly meal plan, priced grocery list and spreadsheet with ChatGPT agentEvidence: Article or interviewBeginner-friendly
ChatGPT agent created five healthy dinner recipes and a grocery list sorted by store section, checked local prices via Instacart data, and exported everything to Google Sheets.
ChatGPT agent · Personal · Source date:
ChatGPT agent was asked to search a real-estate site for 4-bedroom Melbourne homes, save listings and route Saturday inspections, but the site and agent blocked each other.
ChatGPT agent · Personal · Source date:
ChatGPT agent filled an accelerator application using the poster's context and chat memory; the poster rated it 8/10, and also had it research vendors into a spreadsheet (7/10).
ChatGPT agent · Ops · Source date:
- Plan a Budapest weekend and start an Airbnb booking with ChatGPT agentRecreated promptEvidence: Described by poster
ChatGPT agent researched accommodation, activities and vegan restaurants for a Budapest weekend and got into an Airbnb reservation flow, then handed over to the user for login and payment.
ChatGPT agent · Travel · Source date:
- Rename, organize and de-duplicate files with Claude CodeRecreated promptEvidence: Described by poster
Claude Code organized and renamed the poster's files and folders and found and removed duplicates; the poster says it worked incredibly well.
Claude Code · Personal · Source date:
- Request a furniture quote with ChatGPT agentEvidence: Screenshots or logs
In 11 minutes ChatGPT agent ran 13 searches, skipped JavaScript-heavy sites it couldn't read, and shortlisted wicker daybeds from server-rendered retailers so it could request a quote in Sydney.
ChatGPT agent · Shopping · Source date:
- Diagnose a Kubernetes pod-scheduling outage with Claude CodeRecreated promptEvidence: Company case study
Anthropic's Data Infrastructure team gave Claude Code dashboard screenshots, and it traced the problem to pod IP exhaustion and supplied the commands to add a new IP pool; Anthropic says this saved about 20 minutes during an outage.
According to Anthropic:20 minutes saved during an outage
Claude Code · Ops · Source date:
- Fill an online grocery cart from a handwritten list with ChatGPT agentRecreated promptEvidence: Article or interviewPartial
ChatGPT agent transcribed 15 of 16 items from a photographed shopping list (missing onions), found the nearest Harris Teeter and added them to the cart; a safety monitor first blocked the Kroger login, but after the user said they'd log in themselves and entered the password, it ordered the groceries.
ChatGPT agent · Shopping · Source date:
- Three-way match purchase orders against delivery and invoice docs with ChatGPT agentRecreated promptEvidence: Screenshots or logs
In about 4 minutes ChatGPT agent checked that all 5 POs had matching delivery and invoice documents with no variances, combined the PDFs, and wrote a short report.
Poster says:Done in around 4 minutes
ChatGPT agent · Ops · Source date:
ChatGPT agent booked what the poster called the perfect Airbnb on the first try.
ChatGPT agent · Travel · Source date:
- Build an Nvidia DCF valuation spreadsheet with ChatGPT agentRecreated promptEvidence: Screenshots or logsBeginner-friendly
ChatGPT agent ran for 37 minutes, walked through DCF assumptions and returned a downloadable Excel file with formulas and numbers.
ChatGPT agent · Research · Source date:
- Find a family vacation rental in Costa Rica with ChatGPT agentRecreated promptEvidence: Described by poster
ChatGPT agent found a Costa Rica place with a pool at a decent rate for 2 adults and 3 kids; the poster was tempted to book without checking further.
ChatGPT agent · Travel · Source date:
- Build an early-retirement (FIRE) plan with ChatGPT agentRecreated promptEvidence: Shown on videoBeginner-friendly
With early access, the poster had ChatGPT agent spend 20 minutes researching Vancouver tax law, modeling savings and FIRE scenarios, and producing a downloadable presentation.
ChatGPT agent · Research · Source date:
- Find and summarize files in Google Drive with Comet AssistantRecreated promptEvidence: Shown on videoPartial
Comet Assistant kept trying until it found the right files in Google Drive and found a way around opening them, but its summary skipped parts of a PRD after reading only about half.
Perplexity Comet Assistant · Research · Source date:
Claude Code read and changed the local Apple Notes database and organized 225 notes in about 5 minutes with little manual effort.
Poster says:225 notes organized in about 5 minutes
Claude Code · Personal · Source date:
- Order groceries from a 16-item list with OpenAI OperatorRecreated promptEvidence: Article or interview
Operator misread parts of a 16-item grocery list and added only 13 items to the cart even after corrections; the author concluded it wasn't ready for real use.
OpenAI Operator · Shopping · Source date:
- Build a parenting shift-tracker app with Replit AgentEvidence: Screenshots or logs
A non-coder built ParentShift, a working app for couples to track night-duty shifts, using Replit Agent; the author reports it took about 3 hours and cost under $50.
Poster says:Idea to app in 3 hours for $50
Replit Agent · Personal · Source date:
Emily Lambert built an app on an n8n agent that researches a podcast topic, pulls news and tweets, then emails a script/outline to the team.
n8n agent · Content · Source date:
- Turn a blog post into a Quarto slide deck with Claude CodeRecreated promptEvidence: Described by poster
Claude Code read one of the author's blog posts and turned it into a Quarto slide deck in three prompts.
Claude Code · Content · Source date:
Manus studied the author's products and writing, researched SEO and keywords, and produced a content calendar in a 30+ minute run that the author says cost about $10.
Manus · Marketing · Source date:
Security startup Jit moved its engineers from JetBrains to Cursor and, according to the article, delivered features 1.8–3x faster.
Reported by LeadDev:1.8x–3x productivity gains
Cursor · Coding · Source date:
According to Zapier's customer story, Slate Studios produced 2,000+ qualified leads in one month with a single Zapier Agent that also handled data reformatting and competitive research.
According to Zapier:2,000+ leads generated
Zapier Agents · Marketing · Source date:
- Generate a colophon page for a GitHub Pages site with Claude CodeEvidence: Screenshots or logs
Claude Code wrote a script that builds a mobile-friendly colophon page listing every tool and its commit history, then wired it into the GitHub Pages deploy workflow.
Claude Code · Coding · Source date:
- Compile a list of China-tech reporters with ManusRecreated promptEvidence: Article or interviewBeginner-friendly
After being pushed to be thorough, Manus produced a downloadable list of 30 journalists covering China tech with their outlets and notable work, though paywalls and captchas got in its way.
Manus · Research · Source date:
- Vibe-code a 3D town-building game with Claude CodeEvidence: Article or interview
From a single informal prompt, Claude Code built a working 3D game for placing buildings and driving around in about four minutes; follow-up prompts added a fire truck, traffic, and fires.
Reported by One Useful Thing:Working game in about four minutes
Claude Code · Coding · Source date:
- Create an interactive elevator-pitch course with ManusEvidence: Article or interviewBeginner-friendly
Manus researched the topic and built a complete multi-module course website on elevator pitching; a second prompt added interactive quizzes and videos.
Manus · Research · Source date:
According to Zapier's customer story, NisonCo grew weekly leads from 270 to nearly 400 (+48%) and saved about $30,000 a year with an agent that scans news for new businesses in its target industries.
According to Zapier:48% increase in leads
Zapier Agents · Marketing · Source date: