Rishav 🌱

We're planting trees wrong, and it's costing us billions! Up to 80% of planted saplings die because they're in the wrong place. We're fighting climate change with one hand tied behind our back. So we fixed it. 🧵 (1/4)

ADHD multitaskers absolutely printing rn > talk to agent 1 > while you wait for it to respond, talk to agent 2 > while you wait for agent 2 to respond, back to agent 1. if still spinning, talk to agent 3...

indie AI dev anxiety is mostly misdiagnosed it’s not the models moving too fast it’s shipping noisy products, weak systems, zero strategy the quiet builders today will look obvious in hindsight

they can never make me hate you @OpenAI. thanks for codex

Your girlfriend isn’t real bro. Epstein got a copy of her genome from her high school gynecologist that he bribed with 100 BTC. She’s a Mossad clone PLANT to distract you from a historic ascension. Cancel the sushi date tonight and LOCK THE FUCK IN. Horse year is upon us.

Voice is becoming software accents on demand, languages erased mid-call when AI speaks for you, identity turns into a UI choice this is the quiet rewrite of communication. audioconvert.ai/british-accent…

im screaming “sudo rm -rf /“ in every call from now on

gm

voice is becoming software accents on demand, languages erased mid-call when AI speaks for you, identity turns into a UI choice this is the quiet rewrite of communication audioconvert.ai/british-accent…

rocket: abort mission. also rocket: anyway, watch me stick the landing youtube.com/watch?v=1huIM_…

spent the weekend stress testing claude code with plugins. came out with an uncomfortable thought. every time a model gets “more usable”, people say it’s getting dumbed down. but what’s actually happening is constraint is moving from the user to the system. plugins don’t just add capability. they formalize intent. once the system knows what you’re trying to do, it can say no faster, earlier, and more confidently. that *feels* like regression if you’re used to raw, unconstrained output. claude code today feels less like a clever intern and more like a senior engineer with opinions. it won’t brute force your way through bad specs. it pushes back. it asks for structure. it fails loudly when the interface between human intent and machine execution is vague. that’s not weakness. that’s a stress test. the real question isn’t “is the model worse?” it’s “did we lose the illusion of infinite flexibility?” power users are going to feel friction first. casual users will feel magic. this gap will keep widening as tools move from chat to workflows. if you want the old feeling of raw capability, you’ll need better specifications, not a better model.

AGI but it's actually a carbon based model that runs on caffeine and trains in realtime. A little slow token output though

science and tech this week felt like a reminder that progress is mostly about better lenses. JWST confirmed a galaxy at z=14.44. absurdly bright. basically the universe saying “i was busy way earlier than you thought.” every time our instruments improve, cosmic timelines get rewritten. meanwhile, particle accelerators helped read star catalogs hidden under medieval monk scribbles. they erased the text, but physics brought it back. history isn’t lost. it’s just waiting for the right tool. someone built a font that renders cistercian numerals using ligatures. monks invented data-dense notation centuries ago. we’re rediscovering it because we finally care about information compression again. on the software side, custom firmware for old sony minidisc gear is alive and testing. planned obsolescence loses when curiosity is stubborn. and then there’s feynman. still readable. still humbling. proof that great explanations outlive every tool they were written on. pattern is obvious: better hardware doesn’t just move us forward. it reopens doors we thought were closed. cosmos, code, or culture. nothing is final. jwst galaxy paper here if you want your sense of time broken again astro.theoj.org/article/156033… astro.theoj.org/article/156033…

text-only LLMs “seeing” with google lens is the reminder: models aren’t getting smarter. they’re getting better at using tools. intelligence is becoming orchestration. news.ycombinator.com/item?id=469712…

My ADHD ass having 47 creative new ideas at 3 am instead of sleeping

security updates this week feel like a pattern shift infra is automating trust faster than humans can reason about it attackers are moving slower but deeper supply chain, dormant shells the game is no longer speed. it’s visibility blog.prosody.im/2026-letsencry…