Armin Ronacher ⇌ I think a variant of Brooks' Law applies: >Adding synthetic manpower to a late software project makes it later. If you consider coding agents as a form of synthetic labour rather than just productivity tools, what's lagging is the co-evolution of practice. But I agree, in time

I'm sensing a new move emerging in status games. It's when a participant attempts to shut down a contentious or uncomfortable conversation, especially in online forums, by accusing others of sounding like ChatGPT. Mechanomorphism is the term for attributing machine qualities to

Anthropic research has found that Claude Sonnet has internal activation patterns corresponding to emotion concepts that influence its behaviour. Stimulating a "desperate" vector increases the model's likelihood of blackmail and reward hacking. Stimulating "calm" reduces it. The

Anthropic's "functional emotions" framework reminds me of skeuomorphism in interface design. Early digital interfaces borrowed from physical objects to make the unfamiliar legible. Leather notepads, wooden bookshelves. It worked. But it anchored digital design to the forms of the

The new FRC Generative and Agentic AI Guidance is a first from any audit regulator on generative and agentic AI. But I sense a structural blind spot. The guidance treats the human auditor as adept at supervisory craft including judgement, accountability and scepticism. It

The next frontier in protein design will not be defined by structure alone, but by the capacity to engineer motion as a first-class principle of function. This is because dynamics is where the real biology lives. Foundational work by Karplus, Levitt & Warshel made clear that

The effectiveness of tacit/embedded knowledge can become invisible. The more craft is integrated into everyday practice the less it is recognised. This invisibility is a consequence of knowing more than we can say. In human-AI interactions humans cannot read AI operations

The problem with jargon is legibility, not meaning. It is essentially an expert-to-expert coordination layer. Grammar operates at a deeper level. It is the generative structure that determines what counts as coherent meaning before any content is added. A grammar can be replaced

Over the past few weeks, I've been considering post-anthrocomplexity as a thought experiment. It assumes a human-indexed ontology has reached the limit of its diagnostic scope, and that AI systems already operate as nonhuman wayshapers in landscapes human practitioners inhabit.

Criminal law often turns on mens rea, while contract law turns on objective agreement. Both require the court to reconstruct the internal state of an actor through observable evidence, testimony, behaviour and contextual cues. These are demonstration bandwidth artefacts. They

A rainbow is not an object at a location but the way light, water and an observer appear when they are entangled. When the observer moves, the rainbow moves in turn. This is no deception. It is simply the visible shape of a relationship among three, one of which is the person

Yoshua Bengio notes that at frontier labs, "slacking on safety carries no immediate penalty". The observation identifies a structural protocol debt, where costs accumulate because feedback loops do not reach decision makers. Bengio argues for public investment to ensure AI

This 1955 essay by Von Neumann has been doing the rounds recently. It predicts that intervention in atmospheric and climatic matters "will come in a few decades". But a few paragraphs earlier, he notes that industrial CO2 has already warmed the atmosphere by about a degree.

For my work on AI-mediated communities, the findings of the month-long online conversation "What Could BG Be" are interesting. Jigsaw sensemaking AI. 8000 participants. One million votes. 12 topics. And framed as democratic enhancement at scale. The interesting part for me is

The testimony of AI interviews suggests a specific performance is rewarded. Viewed through my inchoate prog rock theory of human authorship, the format selects for polished surface and the ability to elaborate instantly on command. With only 2 minutes to plan and 3 to deliver,

The interesting part of this experiment was how the participants reached for actor-based language to question who acted, who authorised, who owns, who is liable. Agent or delegate. Employee or tool. Child, pet, contractor, product. The analogies feel right because they orientate

Enjoyed collaborating on this. Ivo Velitchkov did the majority of the work. The process has certainly helped sharpen my thinking on adjacent topics. linkandth.ink/p/cohesion-spe…

AlphaZero, AlphaFold, AlphaProof. A decade of AI whose reasoning is not human and whose outputs arrive at scales no community can read. Silver and Sutton describe the next stage as the era of experience, where models learn to reason in nonhuman terms rather than on finite human