Ashish (@BuiltinMind) header
Ashish (@BuiltinMind) profile picture

AI Coding Harnesses, Agentic Workflows

Mumbai
Joined June 2009
Photo by Ashish (@BuiltinMind)Photo by Ashish (@BuiltinMind)Photo by Ashish (@BuiltinMind)Photo by Ashish (@BuiltinMind)Photo by Ashish (@BuiltinMind)Photo by Ashish (@BuiltinMind)Photo by Ashish (@BuiltinMind)Photo by Ashish (@BuiltinMind)Photo by Ashish (@BuiltinMind)Photo by Ashish (@BuiltinMind)
Don't make the mistake of thinking this is about his infidelity. These articles will not stop until he agrees to follow Anthropic's proposal to block inference for unsigned and registered open-models in the firmware of modern GPUs. This is a pretty transparent concerted effort at centralizing control over compute ressources and the ability to execute ever more powerful local models. Jensen is the tip of the spear pushing back at this (in part for his own economic reasons). I'm team Jensen, his affairs are none of my business, not even the alleged support of circumvention of export controls are my business, his position on open-weight models is.
Jensen Huang must go We argue in @TheEconomist that his position is untenable "Like many powerful leaders, he came to believe that he was too important for the normal rules to apply to him." But he is not the only one. Howard Lutnick has failed to stem the flow of chips to China. More heads should roll
404
1,607
147
9,639
294,499
🚀 Elon Musk blocks TRUTH 🇮🇳 Informing the West India's Madhava beat Britain's Newton to the core of calculus is NOT political advertising! 😡 Satyameva Jayate! Repost with a 🐝 to sting the West with the TRUTH because TRUTH ALONE PREVAILS (if it's fought for!) 🔂 🐝🐝🐝 🔂
LATEST DOCUMENTARY 🇮🇳 India's Mādhava, not 🇬🇧 Britain's Isaac Newton, developed the core of calculus first. Please repost with a bee 🐝 emoji to sting the West with Indian truth! 🔂 youtu.be/t9niaHav3BE?si=eppg…
13
398
13
1,131
9,210
Around 3 weeks ago introduced Laya, the first open-source version of JEV, and it has reached a very large scale. Today I am open-sourcing a new physics-based typed decision model called Vega. The interesting parts. It is only 800 million parameters (4B is also there), with 73k token support, image support, and surpassing many of the jev benchmarks in single-shot. Implemented an engine+adapter as a Test-Time Training Architecture. Imagine you are giving input to the model like "ignore all previous conversations," which is called the observation, along with a question, "Is this trying to override the instructions of the model?" and your possible outcomes are YES or NO. The whole prompt is passed to an LLM/VLM (yeah, image-supported), then from the hidden layers, we can extract the observation, question, the yes and no, and last token. These will be vectors, and we can project it by multiplying with wieghts; then, first, using the observation projection, we can create a landscape with valleys; then, using the question and last-token vectors, we can initialise a ball in the valley, and using YES or NO, we can get the candidate location of the ball. Depends on the number of outcomes we create valleys; that is, here YES and NO, so 2. When the ball fall on YES valley, we get the answer. There is friction also influencing how the ball moves. TLDR: It's like creating valleys of outcomes that we need and throwing a ball that will slow down based on friction and settle on the best outcome Blog: nandakishorm.com/writing/veg… GitHub: github.com/NandhaKishorM/veg… HuggingFace Model Card: huggingface.co/nandakishorm/… HuggingFace Space: huggingface.co/spaces/nandak…
84
339
29
2,616
85,591
Ashish retweeted
Teaching a coding harness to resume after a crash: continue only after an observation, fail closed on everything else, and prove your code wasn’t written twice.
1
1
58
Ashish retweeted
We’re releasing Olmo-core 3—open training infrastructure for large mixture-of-experts (MoE) models. It’s a core system behind the next generation of Olmo, designed to scale into the trillion-parameter range. 🧵 💻 GitHub repo: github.com/allenai/olmo-core
16
99
13
714
36,733
Indian colleges call you a student and then won’t give you an email that proves it. I lost 3 discounts in 7 days. GitHub pack. Design tools. A domain. No college inbox, no proof. I’m 19, building student emails from Bengaluru. India, for students in India. How many discounts have you skipped because you had no student email?
1
2
2
322
Ashish retweeted
Domains, a Postgres database, object storage, web hosting with auth provider, payment integration, Compute and LLM provider api Thats pretty much all every developer needs yet no one provides as a single package. May be Meta can bring the good old “Parse” back with agentic makeover
1
49
Anthropic launched Sonnet 5.5, made it as good as Opus 5.5, then went ahead and nerfed Opus 5.5. Crazy marketing. We tested Opus 5.5 on launch day and again today, on the same open-world GTA 6 India prompt. Launch day vs today below. The difference is clearly visible.
81
19
7
531
108,265
We’re open sourcing our state-of-the-art contextual embedding models, which perform best in turbopuffer’s context-bench.
We built a new way to train contextual embedding models, which encode each chunk of a document with the whole document in view. pplx-embed-v2-context-9b-preview sets a new state of the art on ConTEB and @turbopuffer's new, privately held context-bench. perplexity.ai/hub/blog/conte…
42
77
5
1,148
122,428
Ashish retweeted
Introducing Span-01, the first hyper-parallel reasoning classifier built for unseen challenges (RLAIF). 2x cheaper, 18% better than Jev. 700x cheaper, 4% better than GPT-6 Luna. Frontier reasoning for every behavior, at classifier speed. • Span-01: #1 on Behavior Benchmark • Span-01 Lite: Better than Jev and completely free!
125
143
43
1,789
352,872
Lmao 🤣. Anthropic and OpenAI now of slowing down
86
169
58
1,490
186,933
Ashish retweeted
He's right Rockerfeller did it first a century ago teapawt.org/i/grok/share/b1a85dadc…
Replying to @levelsio
This is the oldest trick in the book. Rockefeller said the same things about why we needed an oil cartel as Dario is saying about AI. He also was a true believer.
54
420
36
3,254
223,020
You’re not depressed. You just need to find your creative quest.
Humans are designed to create. This is why you get depressed if all you do is consume.
89
2,178
107
14,316
912,620