Siddhant Minocha

@archiexzzz I’m literally about to board, haha.

Went on a short trip to Goa. We stayed in a villa and a resort, did a lot of water activities, and ate amazing food and drank great drinks. All of that cost just ₹30k (about $300) per person, including flights. It's crazy how cheap you can travel in India.

@rand_longevity @irkedgb but not in the 3-9 months horizon. This one’s a tricky business.

@irkedgb yes

notice how many people in ai are talking about curing all diseases this trend is gonna continue and you are gonna see some major breakthroughs in the next 3-9 months

@OpenRouter Why is it not a surprise.

DeepSeek V4 Flash has topped the weekly leaderboard

@agupta We are now looking at GWAS studies for Aura, Vibes, Chadness and more.

in a reverse uno move, trusting this puts the parent's IQ score in the 00.01th percentile

@zackdfilms1 We just the need of energy of sun for a day.

Spinning Mt Fuji Like A Beyblade 🤯

@reach_vb I do like my VS open to make minor text/css changes. I feel too guilty wasting water to center a div.

realisation: I haven't opened an IDE in more than a month, diff view + file viewer in the codex app is more than sufficient one app to rule them all!

@davidasinclair oversimplification

Aging is a drift away from optimal gene expression

Google’s new quantization work compresses KV cache memory by ~6x and speeds up attention compute up to 8x without retraining. Early signs that TurboQuant may matter for bio AI have started coming in. A new TurboESM paper adapts Google’s TurboQuant to protein language models, showing ~7x KV-cache memory reduction on ESM-2 650M while preserving high embedding similarity. Why this matters: protein/genomic models often hit memory limits before they hit “intelligence” limits. Cheaper long-context inference could mean: • longer protein sequences • larger batch screening • more deployable lab-scale bio LMs • faster embedding search over biological databases Still early, but this is the right kind of systems work for bio AI.

@gokulr If you tell it a human will review your code, it will ship serious crap.

Is it my imagination or is it true that if I tell Codex that the code is from Claude Code (or vice versa), it seems to review the code extra agressively and adversarially?

@buildwithsid @vinniepsychosis OpenRouter isn’t a provider. It’s an aggregator.

@vinniepsychosis use it from trusted providers like openrouter perks of it being opensource

just tried gemini 3.5 flash deepseek v4 flash >> gemini 3.5 flash there's no comparison 💀

@sumanthraman An actual Tsunami in the sea is also coming. It might be an year or two, or maybe 5 or 10. But it is coming.

@growing_daniel wait what

one good girl is worth a thousand GPUs

@sanketnadhani have you tried human?

Back in the market again for a dictation tool. Done with Wispr Flow, done with Monologue, couldn't get Muesli set up. What else?

@silvirouskin haha. my wife often asks me should i be happy or sad?

Me explaining our science to a good friend of mine who’s very smart but a non scientist and without college degree

@amaturefuturist @parmita

@sidwantscoffee @parmita 3 - 6 am is a healthy sleep routine

i will never pull an all nighter again

@isaactolley_ I'm tired of people fake congratulating each other. And if you really mean that, please write on your own. The amount of AI slop, urgghh

anyone else find writing on linkedin a conundrum?

@1Umairshaikh what about the existing 20m who are now having existential crisis?

Everyone thought AI would replace developers. Instead, it created 10 million new people who suddenly think they’re developers.

@weswinder Have started testing it. Yet to run internal benchmarks. I have a gut feeling that I like it more that 3.1 pro. Haven't been able to quantify it yet.

i haven't seen a single good post about gemini flash 3.5 does anybody actually like it?

@chayprabs @weswinder What Sonnet price brother? It's almost half on both input and output.

@weswinder Everyone hates it. Why would someone pay sonnet's price for a flash model.

@davidasinclair That remains largely intact too. The only thing that changes is what it means.

The genome largely stays intact. What changes is how it is read