8 years ago I decided done was better than perfect, and PatternShift was born. And my music and sci-fi writing made it out into the world before the generative AI bomb-pulse.
This Friday, PS dies a fiery, explosive death w/the final album, FRAMESHIFT.
https://t.co/2FFdcP9BZn
@htmx_org@almighty_lisp@LukasHozda lisp & htmx best friends no matter what
been saying this for years*
https://t.co/JwbxvKFpTQ
*at least two, so technically plural
@htmx_org some of the people in your comments have never written a jvm service that's been running on a single instance in aws for years without even a restart handling thousands of requests a second with a buttery smooth response rate and it shows
@adamzwasserman@NinadSachania@htmx_org i mean the right way to do it is to teach them both, so they know which option to use, when, and why, and the tradeoffs of each.
then also teach them not to be twitter zealots
(1) the best way to produce a formal specification of a program is to write a program, & (2) bSAT is the reference NP-complete problem.
formal methods in practice has shown that formal specification takes sig. more work than writing robust software through discovery effort.
Right now the biggest barrier to AI-driven formal methods is that AIs are absolutely dogshit at coming up with good properties. I thought that was just a March 2026 thing but it seems to be a September 2026 thing too
@CosmicEggEarth@steida no. complexity is entropy. you can't win by leveraging it with ai. insight yielding simplicity is still the way.
ai junkies can't touch dominating apps of the past that had first significant cuts built on the timescale of weeks (linux & git, ffmpeg, instagram, craigslist).
you don't need LLMs to write functional code.
- craigslist was built in a few weekends
- core ffmpeg in a few months
- linux to initial version in 5 months and git in a few weeks
- instagram 8 weeks
- stardew valley even @ 4 years destroys LLMs
LLMs suck
the insane amnesia these AI fanbros have; they think the existence of coding agents somehow cancels out the fact that people have built incredible software with solo dev effort in short periods of time w/life-changing consequences.
no, it's not 'early', it's at the saturation point. vibe coding for months is way worse than what people get out of weekend or vacation projects. the only 'early' that makes sense here is 'early' as in 'it will be awhile before it's not way worse than competent humans'
@GohilHardy Honestly it's still early fam. Enterprise stuff needs context and accountability that prompts alone can't give yet. Solo devs will get there though 🚀
yeah totally not made up numbers. totally not highlighting "code" as the metric b/c there's not a lack of meaningful outcome to attach to it otherwise.
but sure let's amplify that implied 80th percentile tier effort so that it drowns out everything better, what could go wrong?