CuriousNerd

An IIT Bombay alumni initiative.

[CuriousNerd @_CuriousNerd]

@nia_thinks Steering the build

[CuriousNerd @_CuriousNerd]

@sanketghatte23 No GPUs for you

[CuriousNerd @_CuriousNerd]

@Adidotdev Well then I'm the home to those.
Please follow to get AI updates 🫡

[CuriousNerd @_CuriousNerd]

@nicdunz Wish I could say the same

[CuriousNerd @_CuriousNerd]

@aryanlabde True, you need to be a better judge

[CuriousNerd @_CuriousNerd]

@divaagurlxw Layoff 🥲

[CuriousNerd @_CuriousNerd]

@markproduct claude

[CuriousNerd @_CuriousNerd]

@kritikakodes Layoff.

[CuriousNerd @_CuriousNerd]

@saidotdev LLMs are just autoregressive models. They predict the next word. So it heavily depends on how well you prompt.
It is not as dumb as we think.
Consider watching this: https://t.co/UFanwUJc0J

[CuriousNerd @_CuriousNerd]

@sudoingX This channel.

[CuriousNerd @_CuriousNerd]

When you try to control the open-source, it won't make you look good @DarioAmodei
What are your views on this?

#ai #anthropic #claude #opensource #viral #opinion

[CuriousNerd @_CuriousNerd]

@nicdunz Monopoly or duopoly is never good.

[CuriousNerd @_CuriousNerd]

https://t.co/j02idG9xOV

#ai #llm #chatgpt #claude #openai #anthropic #grok

[CuriousNerd @_CuriousNerd]

@Aditya_181105 Absolutely!

[CuriousNerd @_CuriousNerd]

5/ JS, TS & Python. 100% local. No API key.

npm i -g @curiousnerd/keel
keel check .
⭐ https://t.co/C5uikWSWdE
v0.2, building in public. If you've ever yelled "I TOLD you we use Zod" at your AI — tell me what it finds. 👇

[CuriousNerd @_CuriousNerd]

Your AI coding agent has amnesia.

It forgets last session's decisions. Writes formatDate three times. Says it uses Zod, then imports Yup in 7 files.

You don't catch it, you didn't write the code.

So I built a tool that does: https://t.co/DMcFmt4p9N

#ai #opensource #llm #codex

[CuriousNerd @_CuriousNerd]

4/ On a real repo it found 41 stale doc refs + 46 duplicate functions in ~2s, for $0.

And the best part: an MCP server. Your agent asks keel before it writes — "does a date formatter already exist?" — and reuses it.

Prevention, not cleanup.

[CuriousNerd @_CuriousNerd]

3/ So keel does the opposite. It doesn't generate context — it verifies it.

Reads your real code, deterministically, for $0, and catches: • drift (docs vs code) • stale doc links • duplicated logic • library conflicts
→ one Coherence Score, 0–100.

[CuriousNerd @_CuriousNerd]

2/ The obvious fix: have an LLM auto-generate a big context file.

Turns out that backfires. ETH Zurich (138 real tasks) found LLM-generated AGENTS.md files lowered success + raised cost.

More context wasn't better. It was worse.

[CuriousNerd @_CuriousNerd]

Well well is it really this good or just benchmaxing?

[Ornith @ornith_]

Aloha! 🌺 Meet Ornith-1.0, a family of open-source LLMs specialized for agentic coding.

Ornith-1.0 spans the full parameter sizes including 9B Dense, 31B Dense, 35B MoE, and 397B MoE. It achieves state-of-the-art performance among open-source models of comparable size on coding benchmarks including:
✅Terminal-Bench 2.1(77.5)
✅SWE-Bench(82.4 on verified, 62.2 on pro, 78.9 on Multilingual)
✅NL2Repo(48.2)
✅SWE Atlas(41.2 on QnA, 42.6 RF, 39.1 TW)
✅ClawEval(77.1)

Post-trained on top of gemma4 and qwen3.5, Ornith-1.0 employs a novel self-improving training strategy in which reinforcement learning is used to generate not only solution rollouts, but also the task-specific scaffolds that drive those rollouts. By jointly optimizing the scaffold and the resulting solution, the model generate higher-quality solutions in agentic coding.😎

All models are released under the MIT license, enabling full commercial and research use.

📖Tech Blog: https://t.co/qT9N2HYWFn
🤗Huggingface: https://t.co/PRrwqjeBtM