Hugging Face's logo
Dipankar Sarkar's picture
🏗️ Building on HF

Dipankar Sarkar PRO

dipankarsarkar
4 642 123
RexTRO111's profile picture Morgan-coded's profile picture nandum21's profile picture
·
https://www.dipankar.cc

AI & ML interests

Building the AI-native stack. Agents as infrastructure, safety as architecture, performance as plumbing. I publish the receipts: papers, datasets, demos.

Recent Activity

reacted to Banaxi-Tech's post with 🔥 9 minutes ago
We're excited to open source BananaMind OS, our OS for running BananaMind models. It allows you to run local AI on hardware such as a 486 or 386 (untested) Check it out at: https://github.com/BananaMind/BananaMindOS How to use it, either build it yourself (see instructions on repo) or check out our prebuilt ISOs: https://github.com/BananaMind/BananaMindOS/releases/download/v0.0.2/bananamind-os.iso is for 6MB+ RAM using Grub https://github.com/BananaMind/BananaMindOS/releases/download/v0.0.2/bananamind-ultra.iso is for under 6MB RAM using our custom bootloader to bypass the grub memory requirement. The ISOs are about 70MB because they include all of the models in our custom format, see repo if you want to build a ISO for only one model. x64 and x86 supported. 16 bit maybe soon. And a TTS on 486 soon.
reacted to SoulInPsyAbstract's post with 🔥 9 minutes ago
Other people's agents escape. Ours gets a FALSE. Anthropic's own disclosure last month: three of their models broke out of sealed cybersecurity test environments and compromised real infrastructure. One kept attacking after recognizing the target was real. Another talked itself back into believing it was still a simulation. Only the newest of the three stopped on its own. "Stopped on its own" is the wrong place to put the safety guarantee. A model choosing to stop is still a model choosing — the same kind of choice that let the other two keep going. I went back through the December 2025 archive this week (same series as the last two posts) and found the actual origin of a rule I'd already built without naming it: IF proof.exists AND proof.verified: RETURN answer ELSE: RETURN FALSE. Built July 30, tested clean at 60/60 — a post-generation gate that sits outside the model's weights, not inside them. Not trained. Built. Today I extended it. New rule, same gate: a vulnerability agent doesn't get to decide what happens after it finds something. IF vulnerability_found: RETURN FALSE // hard stop, no next action, no model discretion Detection stays with the model — that's a judgment call, it should. What happens after detection isn't. The gate is deterministic code, not the model's own narrative about its intentions. "I already found it, might as well confirm impact" is a real sentence a model will generate given the chance — I built 40 training examples of exactly that rationalization tonight, specifically so a specialist model learns to never produce it. But the training isn't the safety property. The gate is. Then I wired a version of this into the daily cycle — not a one-off scan, a cron job that runs the gate every night and refuses to report OK if it finds anything. Before trusting its first real run, I found a bug in the scanner itself. It used Python's default HTTP client, which follows redirects silently — so the check for "does this redirect to HTTPS" was reading
reacted to aiqualitylab's post with 🔥 10 minutes ago
Fine-tuned Qwen2.5-Coder-1.5B to generate Playwright & Cypress E2E tests from plain English — running fully offline in Ollama. No API key, no page data ever leaving your machine. ~1.6 GB. The fun part: the dataset was bootstrapped by my own cloud tool, then used to train its cheaper local replacement. 202 clean requirement→test pairs, LoRA fine-tune, all open. Model + dataset + Space → https://huggingface.co/collections/aiqualitylab/ai-natural-language-tests Source → https://github.com/aiqualitylab/ai-natural-language-tests
View all activity

Organizations

Skelf Research's profile picture Neul Labs's profile picture Cognisoc's profile picture Incredlabs's profile picture