You code in text, then you review in text, and that's pretty ass.
That's why were introducing Alchemize concept graphs, the first visualization that relates concepts and shows how code flows.
Visualize at different depths and jump straight to the code.
You can tell how thin a company is wrapping around foundation models based on how much they care about token costs.
If a company is hella optimizing token cost, they’re probably reselling the intelligence at a low margin, and thus adding very little value
Introducing Agent Prompt Tracking in @TryAlchemize.
Code reviewers can now see the exact prompts behind each line of change, making it easier to distinguish author intent from agent assumptions.
as an early stage founder, "what's your moat" is a meaningless question.
like dude, i've been working on this for 1 month. any competent team can copy the playbook to build and gtm.
early stage, the only moat is if the founder got that dog
(or relationships, legal protection, compounded vertical expertise)
i really enjoy sending/receiving break up messages.
there’s something about getting one that’s deeply thought through. One where you can tell they’ve sat with your history together, confronted what they’re feeling, probably shed a few tears writing it. it hits hard.
conversely, as a writer, it's strangely beautiful too. a moment where you get to sit down and reflect on many moments together. the first coffee date that wouldn't end. your worst moments where they stood by your side and pulled you up. the thorny arguments that just never quite resolved. it's a cathartic blur that (hopefully) ends in you being more confident in your decision and hitting send.
break up messages are one of the few moments where we give ourselves permission to feel everything at once.
Devin and Factory started closest to the endpoint.
Their evolution is:
single agent → parallel fleets → continuous workflows → security/governance → self-improvement