a dev org i just started to work with thought they could just automate testing. bought an AI tool, pointed it at the codebase, cut QA in half. the AI was actually better at catching regression bugs. looked great on paper. but they shipped the buggiest quarter in company history

x.com/i/article/2030…

AI can build a SaaS in a weekend. that's the problem. because most of what gets built in a weekend is a feature, not a business. >> a feature solves one moment of pain. >> a business owns a workflow. when building costs zero, more features get built. revenue stays flat

You have to be able to do it in 5 words or less. If you can’t pitch your thesis concisely, you don’t understand your value prop well enough. Building. Hiring. Fundraising. Investing. Everything.

Autoresearch works even better for optimizing any piece of software. make an auto folder, add program.md and bench script, make a branch and let it rip.

AI distractions are big right now In theory I know what do 1. Pick ONE job role that costs $50k+ 2. Build the automation in a week 3. Price at $15k (still 70% savings for them) 4. Sell BEFORE you build but in practice my technical mind just wants to dive deep into the tech

Everyone with an OpenClaw setup or anything comparable should make sure they install this I've been using qmd since the first week it came out. Grep is great until the kb grows to a certain size. qmd is for that leap

Marc Andreessen ( Marc Andreessen 🇺🇸 ) was right about software eating the world but only halfway. Strip any job to its atoms (surgeon, lawyer, warehouse picker, CEO, developer, therapist etc.) and you will find the same repeating primitive: Decisions and Actions. That's it. Every job,

amazon banned junior engineers from AI code after a 13-hour outage. wrong fix. a senior engineer with the same agent and no review process causes the exact same outage. seniority doesn't help when the system lets unreviewed code hit production. we run AI agents for client

The problem with vector databases are not the thing itself but how they are used. All those chunking strategies are poor way how to prepare data. Using single vector embeddings is poor way to represent and retrieve data. Good late interaction models are the way to go. Rich

Happy Weekend everyone. 1M token context windows are out.

AI is making CEOs delusional

tbh, Nvidia GTC Keynote send off song is a banger

start a company assuming: 1) models will become 10x better 2) the only bottleneck for humans is making as many well informed decisions as fast as possible in a great interface lovable are so impressive, they understood this at gpt 3.5 and the same logic still holds

Farmer pays $5–$8 per cow per month. A New Zealand company puts a solar-powered smart collar on cows. It tracks location 24/7, health, temperature, chewing activity, breeding. Farmer just opens a simple app and draws a line on the map. That line becomes the fence. As cows

New supply chain attack this time for npm axios, the most popular HTTP client library with 300M weekly downloads. Scanning my system I found a use imported from googleworkspace/cli from a few days ago when I was experimenting with gmail/gcal cli. The installed version (luckily)