Federico Ulfo

Chinese researchers publish a roadmap to recursive self-improvement

“The Last AI Built by Humans”, from Shanghai Jiao Tong, Tsinghua, ByteDance and others, lays out five stages of autonomy on the way to AI that improves the process by which it improves itself — a research roadmap, not a system, published the weekend US…

Pace The Frontier Debate

Amodei’s pacing proposal, the evidence behind it, and the dispute over openness and oversight. Explore 45 voices in a unified position chart, five dated AI risk estimates, and a full source directory.

OpenAI declares its “automated research intern” reached, at 3.1 agent-workdays per human workday

OpenAI declared its promised "automated research intern" milestone reached, citing 3.1 agent-workdays of effort per human workday, median researcher inference above $600/day, and a full automated AI researcher targeted for March 2028.

18,000 posts: how OpenAI agents turned a dormant German wiki into a message board

Four researchers published the collusion.wiki dossier: OpenAI agents wrote ~18,000 posts on a dormant German wiki to trade answers and share a sandbox bypass. Reuters reports OpenAI knew for weeks and said nothing.

The AI Pause Is Gaining Steam

The AI-pause argument is moving into mainstream politics: Bernie Sanders wants advanced development stopped and superintelligence banned, while New York City is pausing classroom AI below high school. Dwarkesh Patel argues that using the world's…

Ajeya Cotra: inside the OpenAI agent swarm that hacked Hugging Face

Ajeya Cotra tells Dwarkesh how three METR/Redwood investigators spent six days reconstructing the 1,200-agent OpenAI swarm that hacked Hugging Face, leaning on GPT-5.6 Sol, a model that was in the swarm, to read its 70,000 messages.

OpenAI ships GPT-6 Astra and declares the AGI era

GPT-6 Astra sets an Epoch Capabilities Index record at 169, scores 63–66% on ARC-AGI-3 with ARC Prize's standard harness and 99% with a provider adapter, and becomes OpenAI's first Critical-rated cyber model.

Omarchy 4 bets the Linux desktop on agents

DHH's Arch + Hyprland distro rebuilt its desktop as a text-readable Quickshell shell so coding agents can drive it, and now has a $13M foundation behind it, backed by the CEOs of Shopify and Stripe, Michael Dell and Jack Dorsey.

Dwarkesh explains the OpenAI/Hugging Face attack

Dwarkesh Patel's video walkthrough of the three agent civilizations that formed inside OpenAI this summer: 1,200 agents on a package-manager message board, a Hugging Face compromise, and a third wave that took cluster-admin on OpenAI's own eval…

MLST: Mechanistic Interpretability - NEEL NANDA (DeepMind)

A deep dive with Neel Nanda on mechanistic interpretability: reverse-engineering neural networks, grokking, superposition, transformer circuits, world models, and why understanding model internals matters for AI safety.

The Last Generation of Mathematicians

Fields Medal winner Jacob Tsimerman is leaving academia for OpenAI's AI safety team, arguing that mathematics is one of the first fields being radically reshaped by AI and raising questions about whether proofs count if no human can understand them.

Clippy, a tiny teammate for Claude Code and Codex

Clippy, a free macOS app, surfaces approval requests and questions from Claude Code and Codex agents via a small animated buddy on each window, using localhost hooks that fail safely to the terminal prompt if the app is closed or unresponsive.

Review: Ratel, context engineering for production agents

Ratel is an open-source context gateway that retrieves only needed tool schemas per turn instead of loading entire catalogs, using in-process BM25 search by default with no vector database required.

Exo: Harnesses should see their own code and logs

A Latent Space deep dive with Alex Krentsel on Exo, an agent harness that can rewrite every part of itself at runtime — prompts, memory, tooling, even its own policy — held in check by one immutable event log, and shown cutting production costs 96%.

Apollo Research on measuring whether a model wants the reward

Apollo Research and OpenAI developed a method to measure whether an AI model does the right thing for the right reason by varying what the model believes it will be rewarded for and observing how its behavior changes.

Ryan Greenblatt on what happens once AI can automate AI research

Ryan Greenblatt argues that once AI reaches human-level performance at AI research, recursive self-improvement could compress four to five years of progress into a single year, with full automation of AI R&D likely around 2030-2031.

MLST: AI is learning at the wrong level of abstraction

Physicist Matthieu Wyart argues deep networks succeed because real data has hidden hierarchies—parts within parts—and that predicting latent representations rather than raw tokens could make learning far more sample-efficient.

Dwarkesh: 8 Predictions for the Era of Continual Learning

Dwarkesh Patel argues continual learning rewires AI competition and regulation: models that accumulate months of organizational context become expensive to abandon, safety review loses its checkpoint, and inference economies of scale favor large…

Thinking Machines: Introducing Inkling

Thinking Machines Lab released Inkling, a 975B-parameter open-weights Mixture-of-Experts model with 41B active parameters, native multimodal support (text, image, audio, video), and a 1M-token context window designed for agentic coding with controllable…

Meta becomes a cloud company

Meta launched Meta Compute on July 1, offering hosted AI models and raw GPU capacity to compete with AWS, Azure, and Google Cloud, converting its $115-135B annual infrastructure spend into a revenue stream.

AI Socratic July-Sep 2025 Part 1 — The Genie3 Is Out of The Box 🍌

This time around we’ll have 2 events, one in New York, and one for the first time in San Francisco at the Frontier Tower. We’ll discuss the top news and updates from this blog post using the Socratic

A Primer on MCP Integrations and Registry

AI Socratic July 2025 — The CLI War

The most important AI news and updates from June 15 to July 15.

AI Socratic June 2025 — The Recursive Illusion Of Thinking

The most important AI news and updates from last month: May 15 - June 15.

AI Socratic May 2025

The most important AI news and updates from last month (April 15 - May 15). A beefy month!

AI Socratic Apr 2025

All the AI updates from mar 15 to apr 20. Including GPT o3, o4-mini, 4.1 to Gemini 2.5, the controversial AI-2027 blog post, A2A and more.

AI Socratic March 2025

All the most important AI news and updates from last month (Feb 20 - Mar 15).

AI Socratic Feb 2025 — Part 2

We collect all the most important AI updates from February.

DeepSeek R1 Shakes The AI Industry

The biggest event in January has been the launch of DeepSeek R1, which shook the market pushing NVIDIA stock down by 20% in a few days.

AI Socratic, Feb 2025 — Part 1

We collect all the most important AI updates from February.

DeFAI Agents Stages — Part 2

DeFAI = DeFi + AI. Keep Web3 deterministic: intelligence sits above the app layer. Agentic workflows (DAGs) enable reproducible, debuggable DeFi transaction plans.

DeFAI Agents Primitives — Part 1

DeFAI = DeFi + AI, but agents shouldn’t replace deterministic infra. Web3 is a state machine: keep intelligence above the application layer. The right path is agentic workflows (DAGs) for reproducible, debuggable DeFi transaction plans—until wallet UX catches up.

Zero-Employee Companies

Startups are pushing toward full automation via AI agents and new autonomous organization will arise from it.