Steampunk AI

Weekly Links: Some Tweaks and the 2025 Whirlwind

2025 was a wild ride in AI. In this post we recap the year and there are some updates on the newsletter going forward.

Saturday Links: Icebergs, World Models, and Regulating Emotional Interactions

Job loss studies continue to mount up, regulations for chatbot interactions, world models, and the shape of the AI frontier.

Saturday Links: Agents meet Banks, Watermarking, and Google Pushes Harder

The EU begins to set rules for AI watermarking, Google and Mistral release new models + UK Banks begin adopting Agentic AI.

Sunday Links: Inevitable Jailbreaks, Disney Licensing, and Thinking Machines' Tinker

Disney frees its characters for Sora, OpenAI adopts skills, Tinker goes live, and MCP goes to the Linux Foundation.

Sunday Links: A Snapshot of the AI Future

Sunday Links: A Snapshot of the AI Future with five interviews from today's leaders that outline thinking about what's next.

Sunday Links: Toy Safety, Dogfooding, and Shared Task Representations

Large companies push AI internally (and their own products), AI Safety for toys is a hard problem, and neural structures in monkeys point to shared representations for generalization.

Sunday Links: Group ChatGPT, Deep Learning Weather Models, and American AI Regulation?

OpenAI releases group chats features for a subset of ChatGPT users, Anthropic pushes for AI regulation, and weather models get the deep learning treatment.

Sunday Links: Wikipedia, Microsoft Strategy, and AI Destruction

Google search messes up, Anthropic Claude used in Cyberattacks, Microsoft CEO shares long-term strategy, and the new book predicting AI Destruction.

Saturday Links: Browsers, $1.4T, and Advances in Continuous Learning

Reinforcement learning environments, orbital data centres, browser wars and blocking shopping bots.

Sunday Links: OpenAI safeguards, LLM Introspection, and Bubbles

OpenAI continues to make announcements at an unprecedented rate: changing corporate structure, data on usage, and a policy enforcement model. Anthropic detects signs of introspection in models, and markets keep bubbling along.

How AI Systems Will Talk: APIs, MCP, and A2A meet Agent Languages

AI-to-AI communication is harder to achieve than it looks. This post looks at the intersection of Web APIs, MCP, A2A, and previous generations of Agent Communication. Drawing practical lessons for today's system communications.

Saturday Links: ChatGPT Atlas Jailbroken, AI Music, and Continual Learning

AI browsers, Meta closes off WhatsApp to AI, and advances in AI Music.

Saturday Links: Skills, Job Impact, and Adult ChatGPT

Adult ChatGPT is on its way, Labour market AI effects aren't here yet, and Anthropic launches Skills.

Sunday Links: OpenAI Week, TRM, and Reflection.AI

OpenAI makes a slew of announcements, taking on industry giants in every direction.

Saturday Links: Supabase, AI Personhood, and Vibe Working

Sonnet 4.5, Supabase, AI personhood, and other stories from the week.

Saturday Links: PDDL and Symbolic Planning, GDPEval, and Grabbing Context

Planning in LLMs, efficient attention mechanisms and breakthroughs in time series models.

Saturday Links: AI Chip Shops, The Changing Web, Ray-ban Display Glasses

Lawsuits and traffic flow changes are starting to affect publishers on the web. Meta launches new consumer hardware and AI finds its way into your late-night chippy.

Sunday Links: Service Now Agents, Nondeterminism, and Resilience by Design

Perplexity scraping, Safe Intelligence new Substack, Nondeterminism, Robot Massages.

Sunday Links: High Risk AI, Small Language Models, and AI Certifications

Welcome to another round-up of interesting AI news and some general web news as well. Here they are: * Google Gemini dubbed ‘high risk’ for kids and teens in new safety assessment. Together with OpenAI being sued for user deaths within, this is another story on how LLM controls for safety

Saturday Links: AI Fast Food, Token Explosion, and Context Engineering

From AI Charter Schools to how reasoning models are driving the demand for AI Capacity. Another packed week of stories.

Saturday Links: Agents.MD, $1 Government AI, and SLMs

AI leaders offer government $1 AI, NVIDIA argues for smaller models, and Google shares energy data

Saturday Links: Drug Discovery, AI Safety, and Robot Games

Drug discovery, Chrome, Perplexity, and AI as the interface to the online world.

Sunday Links: Chat GPT-5, Low Drama AI Employees, and Chain of Thought Fragility

GPT-5 takes over the airways, we're looking forward to Cynic mode.

Sunday Links: ChatGPT Chats in Search, Vibe Coding, and AI Tool Using

A huge week for releases and advances. Google and Github both release vibe coding tools. Anthropic integrates with Canva and Figma.

Saturday Links: AI Arms Races, Superintelligence, and AI Raw Dogging

AI economics are unpredictable, but starting to feel inevitable.