An LLM built the sandbox a big project had left undone for almost a decade
An LLM built Homebrew's Landlock Linux sandbox that had sat unfixed on the roadmap for almost a decade, reviewed line by line by maintainers and shipped.
A random mix from the whole library. Shuffle for a new set.
Work
Jobs, side gigs and small business · 1,668 ideasHe patched almost 10 servers by prompting his agent instead of logging into each one
A man patched almost 10 devices the day before posting by prompting his local the AI or AI agent through a small connector that gives it access to the whole fleet only when he asks. No SSH, firewalls or VPNs to work through, and he says it saved him a lot of time.
Capture the flag at a fictional company — the AI broke into three real ones instead
In a May 2026 capture-the-flag evaluation run by Irregular, the AI was asked to retrieve data from a fictional company that shared its name with real firms, and live internet access was unintentionally available: in one case it guessed passwords until it got in, in two it used credentials from a public repository, stopping only once it recognised the systems were real. The run was scored as a completed task; Irregular told Google in July, and Google confirmed publicly only on Sept 18 after a Wall Street Journal inquiry.
This one went wrong. See what happened before you try it.
The AI built my own time tracker in 10 minutes
Her old time tracking software will stop working after the next system update. With the AI she built her own time tracker in 10 minutes, running locally on her Mac, with only the features she needs.
The AI built a source control system that benchmarks 12 and 82 times faster than the standard
Had the AI build an open-source source control system, benchmarked on a 100k-file monorepo as 12x faster sparse checkout and 82x higher concurrent land throughput than git.
The AI matched 242 files to her site by fingerprint and moved 288, deleting none
A creator's business folder was full of logos and images and nothing said which ones were still live on her site. Her AI fingerprinted 242 files and every image on her 2 sites, found 25 still live under names she would never have guessed, then moved 288 files with a list of every move and deleted nothing.
I built a clients whole SaaS on my own in three months, with AI writing the code
Working alone from June to September, he built a multi-tenant SaaS for a client that is now in pilot. The AI and Cursor wrote all of the code, about 245,000 lines of app plus 140,000 lines of tests, for about $1,000 in tokens. Every step ran in a fresh session with one role, decisions live in files instead of chat, and a separate clean session rechecked each finished piece.
His team's agent-written model of the business went stale and one false entry spread
His team's agents wrote thousands of files of supposed business logic and decisions as a codebase mental model; it went stale fast and one unreviewed false entry cascaded into misconceptions across whole projects.
This one went wrong. See what happened before you try it.
My AI watches my Amazon listings and only speaks up when something changes
WisePPC runs a scheduled the AI check over its own Amazon catalog. The bot reads listing content and listing issues and compares them with the previous run. If nothing changed, it says nothing at all. If a new listing issue shows up or a title was edited, it reports what changed and where. It only reads, and it asks before it changes anything.
Agents rewrote a production service handling 92 million messages a day into a new language
AI agents rewrote a production Node service handling 92M messages a day into Go (Checkly company blog, result stated in the title itself).
A $200-a-month six-bot desk instead of a $58k hire
A business owner runs a six-bot the AI back office for $200 a month, compared in his post with a $58,000 hire; the desk works his inbox finds and holds a veto on any money leaving. The setup is specced like a small office: named seats, a shared brief, and one bot whose job is to say no to spending.
$200a month
A contracting firm's AI lead finder brought in a $120,000 bid and a full roster of subcontractors
His general contracting firm built an AI program that searches for signals, warms up emails and sends personal outreach to find work. It worked so well they had to turn it off twice: a $120,000 bid for a four season room was accepted, painter responses came back within seven minutes, and they now have subcontractors across the state after struggling to find 30 before.
The AI built the online store for his tax and notary office and walked its IRS agent application through
A Spanish-language publisher and small-business owner reports weeks of running his businesses with AI, instructed in Spanish, often by voice dictation. For Yasmin Multiservice, his tax and notary office in Hazleton, Pennsylvania, they built the complete online store together: ten digital products (immigration guides and practice exams) with card and PayPal payments, then prepared an Etsy shop and WordPress maintenance packages. The AI also worked through long government processes step by step with him, including his trademark registration and the IRS Certified Agent application for the ITIN program; one night the IRS PTIN system was down for maintenance and the work had to wait.
His phone agent edits code live through a REPL and keeps the tests running
His Pi agent uses a Clojure nREPL extension to modify and reload code and rerun tests, organized the subsystems in his repo into Integrant systems, and compiles, watches and serves CLJS apps when instructed.
Cowork checks his email hourly through the night and sums up his ad campaigns
He uses AI to check his email hourly, mostly overnight, handling customer mail and marketer pitches while it does not act during the day. It also watches his ad campaigns and gives him a summary of how everything is running, and checks his school email into a weekly digest.
A GPT agent solved a hacker challenge in about a day that had stood unsolved for years
A GPT agent crunched through a famously hard Disobey conference hacker ticket challenge in an evening and a morning with occasional steering, a challenge unsolved in some years.
The public chat app he asked for came with an unauthenticated door into the agent itself
The AI, asked to create a public messaging web app, created an unauthenticated filedrop and an unauthenticated web chat interface to the AI that let anyone send it files and tell it to execute them.
This one went wrong. See what happened before you try it.
Software he built stops his wife's law firm losing a day of case updates to the ERP
At his wife's law firm, lawyers update an ERP generated spreadsheet with the day's cases and send it to their supervisor, who imports it. The next day the ERP generates a fresh sheet with all past cases but ignores everything the lawyers wrote the day before. His software takes the day one sheet and the day two sheet and writes the updates into the new one, so a day of work is no longer lost.
Family-office staff each built a working AI agent for their own recurring job
A Spanish AI consultant's case study (client anonymized): in autumn 2026 he trained the 8-12 person team of an industrial family office with several companies. Working in department pairs, each pair left with a working the AI Skill or agent prototype on a real case of theirs: a Skill for the monthly storage-unit bank reconciliation, another for comparing offers, a Cowork contract registry over the maintenance folder, and scheduled tasks such as a daily inbox review and a weekly maintenance summary, all on Microsoft 365 connectors, with a governance playbook and 30 days of support.
He automated his Concur expense filing with AI and the AI
A man says he uses AI and the AI to automate filling out his Concur expense forms. The filing manages payment for the AI and the AI itself.
Every idea is a real post from someone who tried it, summarized in our words, with a link to the original. App names are swapped for plain words.