teo retweetledi
teo
591 posts
teo retweetledi
Really proud of the progress we’re making at @RadicalNumerics.
AI has made incredible strides in the world of bits. But science lives in the world of atoms — messy, complex, and grounded in physical reality.
Bridging that gap requires a fundamentally different approach to models, data, and systems.
That’s the direction we’re pushing as a team, and it’s exciting to see it come together.
We’re growing fast — if you want to work on some of the hardest and most meaningful problems in AI, we’d love to talk.
Michael Poli@MichaelPoli6
We're growing rapidly at @RadicalNumerics and scaling our core teams. Join us in building the next generation of scientific world models. We're hiring across a few roles, each with significant ownership and cross-functional scope: - Member of Technical Staff, Post-Training - Member of Technical Staff, Infrastructure and Training Systems - Member of Technical Staff, Pretraining Science - Member of Technical Staff, AI Bio - Member of Technical Staff, Biosecurity Our technology brings together numerics, systems engineering, and architecture design to tackle large-scale pretraining on scientific data. Our blogs (see below) give a flavor of the work. We believe that advancing capabilities must go hand-in-hand with advancing safety and biosecurity. The same systems that design biology must also help defend against it. Ping me or others in the team if you'd like to learn more. job-boards.greenhouse.io/radicalnumerics
English
teo retweetledi
We're growing rapidly at @RadicalNumerics and scaling our core teams. Join us in building the next generation of scientific world models.
We're hiring across a few roles, each with significant ownership and cross-functional scope:
- Member of Technical Staff, Post-Training
- Member of Technical Staff, Infrastructure and Training Systems
- Member of Technical Staff, Pretraining Science
- Member of Technical Staff, AI Bio
- Member of Technical Staff, Biosecurity
Our technology brings together numerics, systems engineering, and architecture design to tackle large-scale pretraining on scientific data. Our blogs (see below) give a flavor of the work.
We believe that advancing capabilities must go hand-in-hand with advancing safety and biosecurity. The same systems that design biology must also help defend against it.
Ping me or others in the team if you'd like to learn more.
job-boards.greenhouse.io/radicalnumerics
English
teo retweetledi
Our developer conference Code with Claude returns this spring, this time in San Francisco, London, and Tokyo.
Join us for a full day of workshops, demos, and 1:1 office hours with teams behind Claude.
Register to watch from anywhere or apply to attend: claude.com/code-with-clau…
English
えええええええええええ！！！！！！！！！！
OpenAI Newsroom@OpenAINewsroom
We've reached an agreement to acquire Astral. After we close, OpenAI plans for @astral_sh to join our Codex team, with a continued focus on building great tools and advancing the shared mission of making developers more productive. openai.com/index/openai-t…
日本語
teo retweetledi
Excited to release PostTrainBench v1.0!
This benchmark evaluates the ability of frontier AI agents to post-train language models in a simplified setting.
We believe this is a first step toward tracking progress in recursive self-improvement 🧵:
English
teo retweetledi
🚀 W&B Skills for Coding Agentをリリースしました！
スキルをインストールして、数千件のトレースやRL学習ランを含むライブのファインチューニングプロジェクトに接続。そのままターミナルから数秒でクエリできます。
さらに、@weave_wb とも連携可能です! エージェントのトレースを取得し、失敗ポイントを素早く特定できます。
npx skills add wandb/skills
Claude Code、Codex、その他のコーディングエージェントに対応しています！
↯
github.com/wandb/skills
日本語
teo retweetledi
先週はW&Bのハッカソンを開催しました！
- Coding AgentをWeaveでTrace
- Skillsの性能をWeaveで自動評価
- Finetuning用の所望のデータ作成を行うフローの構築
- 人手評価にアラインするLLM as a judgeのプロンプトの自動生成
- Claude codeの待ち時間にAI newsを確認できるツール
などなど、、W&Bのハッカソンらしい継続的に利用できるレベルの高いLLMOps Toolの開発が繰り広げられました！
当日のソリューションは公開できる範囲で一部ブログ化されます！お楽しみに！
#wandb #hackathon #ハッカソン
日本語
teo retweetledi
言語処理学会第32回年次大会 #NLP2026 にWeights & Biases も出展しています！
3/9–12の4日間、P会場にいますのでぜひお気軽にお立ち寄りください 👋
W&Bの新しいホワイトペーパー「AI Agent アプリケーション開発を支えるLLM Observability Tool - Weave」もご案内しています！
日本語
teo retweetledi
Today, we release our largest LFM2 model: LFM2-24B-A2B 🐘
> 24B total parameters
> 2.3B active per token
> Built on our hybrid, hardware-aware LFM2 architecture
It combines LFM2’s fast, memory-efficient design with a Mixture of Experts setup, so only 2.3B parameters activate each run.
The result: best-in-class efficiency, fast edge inference, and predictable log-linear scaling all in a 32GB, 2B-active MoE footprint.
🧵
English
teo retweetledi
ハッカソンの説明会2回分が終了しました！
最後の説明会（全部同じ内容ですよ）は2/27 に開催予定です！
説明会参加時には、正式なハッカソン参加登録は必要ないです
是非ご参加ください
wandb.connpass.com/event/384354/
日本語
teo retweetledi
Great application of skills: post-train models directly from Claude Code or Codex. No code, no GPU required.
We fine-tuned LFM2.5-1.2B-Instruct with just a prompt 👀
Ben Burtenshaw@ben_burtenshaw
You can finetune AI models with Unsloth + Hugging Face, and right now it’s free! We're giving away GPU credits on HF Jobs + Unsloth. Train a 1.2B model like @liquidai LFM2.5-1.2B with one command. Ship it to your phone, laptop, or API. No infra. No setup. Just prompt your coding agent and go.
English
teo retweetledi
This was one of the key learnings of my past week in SF! Great article
Harrison Chase@hwchase17
English
teo retweetledi
This is a big issue we have when trying things out, calling them "failures" where they are actually "learnings" and often blueprints/artifacts that are reusable for the big hit. The power of compounding plays out naturally if we follow that mindset.
Peter Steinberger 🦞@steipete
The funniest take is that I "failed" 43 times when people look at my GitHub repos and projects. Uhmm... no? Most of these are part of @openclaw, I had to build an army to make it useful. github.com/steipete/
English
teo retweetledi
A thought-provoking and grounded manifesto for AI-era engineering. Focus on outcomes, not on the how o16g.com
English
teo retweetledi
W&Bがハッカソンを企画しました！
Vibe Coding時代だからこそ、本当に使われ続けるAgent開発に向き合うハッカソンを開催します。
✏️テーマ
AI Agentの「使われ続ける仕組み」や「品質向上のためのツール」の開発
例）
・生成AIアプリケーションの評価・分析・改善を自動化するAgent
・Vibe Codingの質を上げるために、うまくいくプロンプト／Skillsを自動抽出
・etc.
👉 上記ツール開発のハードルが高い場合は、業務で“実際に使われ続ける”生成AIツールやその仕組みづくりでもOK！
例）
・継続的にカスタマイズされるメール返信Agent
・etc.
📅 2月 開発開始→ 3/7 最終発表
🏆 賞金 1位：15万円, 2位：7万円, 特別賞（Good pain point sharing賞）：5万円 ＋ 各賞 W&B特製スワグ
💰 参加無料
👇 詳細・参加登録
site.wandb.ai/resources/even…
是非ご登録ください！
#AIエージェント #生成AI #Hackathon #Weave #LLM
日本語
teo retweetledi
teo retweetledi
Lots of interest around small models! It was a pleasure ranting about AI and talking about your use cases.
Thanks to @josephinePqt1 and the AI Circle for inviting me!
Josephine Parquet@josephinePqt1
Full room last night for our chat with @maximelabonne! 🐐 "Contrarian opinion": Start with evals - they force you to confront edge cases early and align with stakeholders when it's still fixable, not after you've already built the wrong thing. Thanks @wandb for hosting us!!
English
teo retweetledi
こんにちは。木村です。久しぶりに Mercari Gears へ投稿させていただきました。
これまでマイクロサービス化を推進し、現在は国際展開におけるEngineeringの責任者を担当しているdeeeetさん とともに、メルカリのアーキテクチャのこれまでの歴史を振り返りつつ、うまくいった点や直面してきた課題について深掘りしました。
また、今後メルカリが国際展開を進め、新規事業を生み出していくうえで、どのようなアーキテクチャを目指していくべきかについても議論しています。
「ベストなアーキテクチャ」は一つではなく、会社やサービスのフェーズ、組織規模によって選択されるべきものです。AIの進化が著しい今だからこそ、依然としてエンジニアが真剣に向き合うべき重要なテーマだと感じています。１時間と少し長いのですが、よかったらご視聴ください！
本当は、アーキテクチャを整えたうえでのAI活用についても語りたかったのですが、今回は時間切れであまり触れられませんでした。そのあたりは、ぜひ次回をご期待ください！
youtube.com/watch?v=SC03y4…
YouTube
日本語