Hemendra Shekhawat

C Code to Assembly.

If you don't like opencode, we can't be friends.

opencode go gives you all of these for $10 a month

come work with us!

I am implementing and training diffusion language model from scratch. The challenge is to not eat my favourite food until I figure out obtaining a decent inference performance! > implementing training infrastructure and scaling the training. > I am optimising KV cache performance to speed up the transformers block > Implementing Diffusion Blocks which can predict tokens parallel! lessgoo!

GPT-5.5 Pro Moto GP game in a single script. One of the better results I've ever seen.

deepseek is outrageously cheap even if you ignore the current 75% discount input token prices are 35x cheaper than opus cached tokens are 178x cheaper

theo gets a lot of hate but he was one of the first people to seriously bet on me i built create-t3-app as a random kid on the internet. he gave me my first internet money, distribution, and a network that changed my life cool to see him keep doing this. talent compounds fast when someone backs it early

i just wanted custom cabinets this escalated quickly...

Just dropped two open-source models: MiMo-V2.5-Pro (Code Agent, 1T total) and MiMo-V2.5 (Multimodal Agent, 310B total). Oh and one more thing — we're giving devs & creators 100T tokens on us. Go build something cool 🛠️ 🎁 100T Free Token Grant for Builders 100t.xiaomimimo.com

Now in 𝚘𝚙𝚎𝚗𝚌𝚘𝚍𝚎: Automagical Zed support. @opencode 🫶 @zeddotdev

The Newton–Schulz iteration coefficients optimized by DeepSeek-V4 are surprisingly strong: they effectively normalize all singular values to 1. This matches our previous intuition: a well-balanced spectrum may help strike a better balance across long-tail knowledge. Plot code: github.com/FengzhuoZhang/…

Why does Muon outperform Adam—and how? 🚀Answer: Muon Outperforms Adam in Tail-End Associative Memory Learning Three Key Findings: > Associative memory parameters are the main beneficiaries of Muon, compared to Adam. > Muon yields more isotropic weights than Adam. > In heavy-tailed tasks, Muon significantly improves tail-class learning compared to Adam. Paper Link: arxiv.org/pdf/2509.26030 A thread 🧵

🚀 DeepSeek-V4 Preview is officially live & open-sourced! Welcome to the era of cost-effective 1M context length. 🔹 DeepSeek-V4-Pro: 1.6T total / 49B active params. Performance rivaling the world's top closed-source models. 🔹 DeepSeek-V4-Flash: 284B total / 13B active params. Your fast, efficient, and economical choice. Try it now at chat.deepseek.com via Expert Mode / Instant Mode. API is updated & available today! 📄 Tech Report: huggingface.co/deepseek-ai/De… 🤗 Open Weights: huggingface.co/collections/de… 1/n

no more a raycast vehicle but a shapecast vehicle, could have been a WR but failed miserably at the end #threejs #gamedev

@IAmManware Copilot is decent, I have been using it for past 3 weeks and I am very happy of the interface that it provides and the result it produces. It just overthinks sometimes.

Hey guys, my company is pushing copilot down my throat for productivity. Guess I am a vibe coder from now. Pls don't be harsh on this beginner 🥹 Ps: wtf are these new copilot limits 😾🤮

2nd place at @opencode Buildathon 🥈 @BansalRishit and I built openflip.io - a pocket AI red teaming agent that scans nearby signals (Sub-1GHz, Bluetooth, WiFi, NFC, RFID, IR), finds/generate the right exploit modules via @opencode, and runs them in real time. Flipper Zero… but it thinks for itself. 👉 Join the waitlist: openflip.io Huge shoutout to @GrowthX_Club, @opencode, @udayan_w, @nexxel and the entire team for hosting such an incredible event. Second win this weekend - we are working on building and shipping both these projects in the coming days! Stay tuned.💪

Normal day: Codex again wrote defensive code inside hot inner loops. All of the data is already validated when the object enters the data structure. Super important to always check AI written code and ask it to clean up all the mess. Otherwise technical debt increases gradually.

if you're not using tanstack, start now!