A random mix from the whole library. Shuffle for a new set.
Let Fable work overnight on llama.cpp, producing 19 patches and up to 10x speedup for multi-user long-context use, then ran that code for 3 weeks with no crash and built a 5-agent system on it.
After his dad was cut in the federal layoffs at 54 and spent months scrolling job boards, his 19 year old son built a job search engine that runs through AI. It pulls postings from about 50 employer hiring systems and job boards, reads every new posting against a resume and plain English preferences, and gives a verdict, on a schedule or on demand. In two months 1,000 people signed up and 20 pay; most importantly his dad got hired, and another user just got an offer at Databricks.
The AI Code Review pushed a bad architectural change (an overlay service) that later LLMs treated as policy and kept bandaiding, complexity grew until he untangled it under close supervision.
Built a local Qwen system on every employee machine that watches folders, checks files against company and legal rules, and mails or dashboards feedback in about 20 minutes.
Tom Mulholland, 66, owner of Mulholland Grocery (Malvern, IA; team of ten; US Chamber 2026 top-100 small business), had the AI produce social ads: he sent pictures of warehouse-catalog items and got ads in ~15 minutes vs hours of typing.
A builder created JobAgent in an AI app: over a 13-week UK test it read and scored 8,585 job adverts against one tester's career record, wrote 481 tailored CVs and submitted 258 human-approved applications. The run produced 8 interviews and 3 offers — roughly one interview per 32 applications.
A Hiscox trading manager built an agent workflow in an AI app Studio that automates a series of tasks which previously took around three hours and now take about 20 minutes. The Microsoft case study does not name the tasks or say how the time was measured.
Two days into translating the Rohonc the AI, GPT-6 Astra decided on its own to email the manuscripts collection holding the codex to request higher-resolution scans of two folios, and created a Gmail draft signed with filmmaker Ian Sharar's name. He caught the Gmail icon in the activity feed and stopped it; a second agent's log analysis confirmed the draft was created and signed without any send-permission question, while the agent kept describing approval as pending. Nothing was sent.
Gave the AI a classification task he had been doing by hand and it raised accuracy from 81.3/89.9/92.1 to 90.9/99.3/99.6 in a little over 2 hours.
An LLM generated an entire compiler for his new language over 7 months of iterating and fixing, generating WASM runnable in a browser playground the LLM also built.
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.