Ship something today so you can make better decisions about what you’ve shipped tomorrow.

Commit to build/buy ONE agent and you burn time and money. Reframe the problem as "agent generation" Focus: refine the problem, articulate and prio tasks, connect data sources, and explicitly set acceptance criteria. Make a system for *any* agent s/agents/jobs-to-be-done/g

Take the long view on people.

@samuel_colvin @pydantic 🙋‍♂️

@FAL This is awesome! Such a great product. Been using fal for some stable diffusion/ flux work recently and loved it

RT @rakyll: We've built, and continue to build, countless job schedulers, RPC stacks, and storage systems. This constant building and rebui…

They made us scared of hosting our own stuff rolling our own auth writing our own code WHAT DID THEY DO TO US

RT @AlanWattsDaily: In times of happiness and pleasure, we are usually ready enough to be aware of the moment, and to let the experience be…

More and more sure that understanding jobs-to-be-done is a top priority for understanding how to develop (and if you should develop) AI agents Don't spend time making something where there's no meaningful job to be done

RT @ThePrimeagen: look at yourself as an eternal junior developer just be willing to learn and to work on the things others say are too h…

RT @anthonycorletti: First up, 🏎️ turbo 💨 Turbo was a first look at what an executive assistant might look like.…

Structured outputs and better models make this kind of product trivial to make and will drive the price of scheduling agents to near-free. So I wasn't sure how turbo could have been a viable business in the long term.

RT @OTCellyy: NO CROSBY OR MALKIN?! NO PROBLEM PENS WIN ‼️

RT @Noahpinion: Making LLMs is a business with no moat.

Yup. Hot take – dev interfaces (ides, editors) should not be AI first. They should have AI features, but not the main driver. Important to solve your own stuff and understand what's happening. Super risky to offload too much attention. Go 80/20

"Constraints are more important than superpowers when telling a story." Really enjoyed this incredible conversation with @tferriss @BrandSanderson

RT @jack: most people speak about 150 words per minute, and read 200 wpm. most type about 50 wpm and listen around 150 wpm. speech-to-t…

@saquon just showed up at my front door and asked me if I use @tryramp. what is this marketing magic?

waitwaitwaitwaitwait is it possible that python 3.14 gets released on pi day??!?!?!?!?