Agentic Engineering | OpenSkills Creator | OSS

Joined February 2023
857 Photos and videos
Pinned Tweet

5
5
38
22,754
TIL the Codex remote feature works even if it’s not your account if you use SSH login Super useful if you’re using multiple accounts or like me, assisting friends and family Setup Tailscale, login with the device password and bobs your uncle
1
7
1,079
Numman Ali retweeted
Jun 4
Anthropicの再帰的自己改善(RSI)についてのブログの中で、Anthropic社内で起こってる話が面白い👀 • 2026年5月時点で、マージされたコードの80%以上がClaudeによって書かれている • エンジニア1人あたりのマージされたコード量が、2024年比で8倍になった • コードの品質はすでに人間と同等レベルに達しており、1年以内に人間を上回ると予想 • コード最適化の実験で、52倍の高速化を達成(Claude Mythos Preview) • オープンエンドなコーディング/エンジニアリングタスクでの成功率が76%に到達 • 一部のエンジニアは「もう自分でコードを書いていない」と公言している このままAIが自分自身の後継を自律的に設計・開発できる状態が進んだ場合のシナリオとしては3つのパターンが言及されている: 1. トレンドが停滞するが、現在の能力が広く普及する 
S字カーブで成長が鈍化する可能性(新しいアーキテクチャの限界、計算資源・電力の制約など)。それでも「100人の会社が1,000人分の仕事をこなせるようになる」ような変化は起きる。 2. 効率向上が続き、AI開発がさらに自動化される 
人間は方向性と最終判断を担い、AIが大部分をやる状態。生産性が劇的に上がる一方で、権威主義的な監視や個人向け操作などの悪用リスクも高まる。100人の会社が1万〜10万人規模の仕事をこなせる 3. 本格的な再帰的自己改善が始まる 
AIが自分で後継モデルを設計・訓練するようになる。ペースは計算資源次第。人間は主に「監督・検証・安全確認」の役割にシフトするが、価値観のずれが複利的に拡大するリスクも指摘されている。 anthropic.com/institute/recu…
Our internal data shows Claude is accelerating AI development—a possible path to recursive self-improvement, or AI autonomously building a more capable successor. It’s happening faster than we thought, and the implications deserve greater attention. anthropic.com/institute/recu…
5
64
284
76,696
Numman Ali retweeted
Our internal data shows Claude is accelerating AI development—a possible path to recursive self-improvement, or AI autonomously building a more capable successor. It’s happening faster than we thought, and the implications deserve greater attention. anthropic.com/institute/recu…
1,771
4,661
28,649
18,494,449
Cool project to generate synthetic agent traces
Today I'm launching a new project called SynthTraces 🔥 It is a minimal codebase to generate synthetic coding agent session traces using Pi (from @badlogicgames) I wanted a large number of coding-agent traces, so I built a tiny harness where two models talk to each other: - an open model (served via HF Inference Providers) plays the coding agent. It gets read bash access to a real open source codebase (the huggingface OSS projects) - a small local model (llama.cpp) plays the human user, asking simple questions like "how do I run this?" or "how is CI set up?" The result is more than 2,000 Pi session traces which can be used to train or fine-tune LLMs, and optimize them for Pi 🤯 And ofc everything is published on @huggingface
7
1,680
Going deep on flueframework.com - The Agent Harness Framework Claude Code Dynamic Workflows are really good for large structured implementations I actually like using the Claude Code desktop app with it, the visualisation is quite nice
1
1
2
999
Codex / ChatGPT design system is become very inviting It makes me feel like the future is mind with the subtle glows and gel like icons Would love to know how they did the design research
3
1
14
1,083
This is super cool, your own company Lovable
Jun 2
Building apps has never been easier. With Sites, Codex can turn your work, ideas, and plans into an interactive website or app your team can explore, use, and share with a URL. Rolling out to Business and Enterprise plans, before expanding more broadly.
3
849
SSH between old and new laptop Codex handling configuration of all CLI tools and configs in one swoop It’s the little things
4
16
2,312
Yoooooooooooooo
Jun 1
Composer 2.5 is now available inside Grok Build. Composer 2.5 is a fast, highly intelligent model that excels on long-running tasks and following complex instructions.
8
1,346
Kind of criminal that @pontusab doesn’t have more GitHub sponsors If you want to support a healthy OSS ecosystem building in the AI Agent space, then add yourself to the list List of projects in thread🧵 github.com/sponsors/pontusab
3
9
8,620
CalText iMessage calorie tracking assistant powered by AI. github.com/pontusab/caltext
1
1
260
Pontus will you leave anything for the rest of us? Too cool project, this is what you’d expect from a frontier lab building iMessage / WhatsApp agents He’s giving it out for free! Definitely going to make a weekend project on this
I just shipped message-ui, build dynamic iMessage attachments with React. ◆ Charts ◆ Tables ◆ Text primitives ◆ Local preview PNG export ◆ Tailwind support ◆ Works with Chat SDK Link ⬇️🧵
543
Codex App > Codex CLI You (and the team!) did it @ajambrosino I want to know how you fixed the performance for multiple agents and long threads A month ago it couldn't handle my workflows Blog post soon?
5
7
2,328
My Claude Code usage has gone from very little to many hours a day thanks to Dynamic Workflows But guess what - this means RLMs (Recursive Language Models) are directionally correct Time to spend time with aithy.dev/ and axllm.dev/ from @dosco

2
1
32
7,970
OSS Calorie Tracking through iMessage Coolest part is that this is the perfect blueprint for making an iMessage assistant, it utilises the Chat SDK under the hood Recommend reading the code (Pontus writes top 1% of code IMO) and fork the repo to experiment with your own ideas
Introducing Caltext, open-source calorie tracking in iMessage. Built with: • Bun Turborepo • Hono on Nitro (Vercel) • Chat SDK Sendblue • AI SDK GPT-4.1 vision • Upstash Redis • Vercel Workflows • USDA FoodData Central Link ⬇️🧵
2
20
3,627