#ai (123 links)
- I've gone back and forth between "we're so cooked" and "we're so back"... news.ycombinator.com
There will emerge a market for code written by humans. This has to happen: I can't prove it, but I can feel it. Despite vast improvements in the generative models, AI-generated art and music are still slop. AI-generated prose has gotten sloppier with phenomena like the much-mocked "Claudese". It's so obviously devoid of a human mind behind it that Zoomers and Alphennials are starting to use "that's so AI" as an insult. Despite the leaps-and-bounds improvements of recent models, I have to believe that AI code is slop too; the major difference between code and these other expressive forms is that we are far more tolerant of slop code: for the most part we just care if the damn thing works.
- Effective altruism is this century’s biggest idea economist.com
It is right to revere reason. But moral philosophers from Aristotle to Rawls knew that if an argument leads to intolerable places, it must have a hidden faulty step. EAs’ instinct, though, is to enjoy counterintuitive conclusions, even if they seem perverse.
- SKF advert review – let’s hope this dull AI Greta Garbo is not the future | The Guardian theguardian.com
The AI Greta Garbo is a bizarre soignée figure, a Stepford robo-star, a Hallmark-greeting-card parody of the icy hauteur, distraite sensuality and deadpan comedy that Garbo projected in life. The AI Garbo is like the classiest ever presenter of a TV shopping channel, purring her way through a dull and humourless script that seeks to create a brand synergy between the frictionless ball-bearings manufactured by SKF and what the company imagines to be Garbo’s own all-important smoothness.
- 2026 in LLMs (so far) simonwillison.net
On Friday I gave the closing keynote at the WeAreDevelopers World Congress North America in San Jose. I tied together the key trends from the past year into a chronological exploration of everything that happened in 2026. The video is on YouTube; here are my annotated slides and notes to accompany the talk.
- Don’t let AI kill the author economist.com
- What Even Is An OS Now? — A Final Ward sockpuppet.org
We’re building a phone that isn’t designed to run fixed-function applications. I doubt anyone will ever get up on a stage and show off a AAA mobile game (that nobody plays) on it. It’s not for TikTok and short-form video, though I guess you can ask it for that. You can ask it for whatever: it’ll build apps for you, on the phone, whenever you want.
- Why I Changed My Mind About AI Risk - by Francis Fukuyama persuasion.community
The primary science fiction scenario for out-of-control AIs involves machines developing their own malevolent intentions. This has provoked a host of discussions as to whether an AI could develop consciousness or have something like human intentionality. These speculations are misplaced, however: an AI does not need either consciousness or autonomous intentionality to become dangerous. It can be directed by a human to achieve one goal and develop intentions to achieve the subordinate goals necessary to carry out the human directive it was given.
- The bastards coming to ruin your business alexmhsmith.substack.com
What an absolute frontier operator.
The cost of writing code collapsed, and the cost of reviewing, fixing and operating it is following, and I'm assuming it gets there. What's left of making software is finding out what people actually want, defining it precisely, and making it pleasant to use. That cost is per piece of software and doesn't transfer, so as the amount of software goes to infinity, which it will because there's no ceiling on demand, that cost becomes the whole job.
- Temperature Zero for Culture: Why Everything Is Starting to Look the Same laurenleek.substack.com
A world that keeps removing the unlikely also removes the things that could have become the next likely. Ecologists have known this for a long time already and even priced it: they call it the insurance value of biodiversity, the standing reserve of rare species that cost energy and contribute nothing, until the climate shifts and one of them turns out to be what survives. A monoculture is efficient right up until the weather changes.
- Why do we keep lying about growth? alexmhsmith.substack.com
Great mini-essay on business and market growth.
AI is the best thing that ever happened to a company like ours. Not because it changes what we believe, but because it stops us having to compromise. The reason we couldn't stay small, flat and senior at enterprise scale was that making at that scale needs bodies. Now, it doesn’t.
- What AI has in common with dogs economist.com
Back when dog technology was first being developed, it was far from obvious that it was safe for people—who were literally inviting wolves into their caves. They had to train them not to eat babies. Thousands of years later, dogs still sometimes escape and go rogue. But overall, people are happy with how things have worked out. (It did not end so well for the Neanderthals.)
Great analogy.
- LukeW | Little Bits of Intelligence lukew.com
Big AI models like ChatGPT and Claude get all the attention, but the continually increasing amount of fast, low cost models means today's software can embed little bits of intelligence anywhere they're needed. And when software can think, we can rethink the UI of software.
- Doorman fallacy - A quick explanation jaakkoj.com
You can think of the doorman fallacy as an extension or example of the Chesterton's fence fallacy: getting rid of the fence (doorman) without fully understanding why it was there in the first place.
- UI components for AI agents aicss.dev
Everything an agent shows mid-conversation: thinking states, tool calls, streaming text, tables and more.
- OpenCode | The open source AI coding agent opencode.ai
- An African vision of artificial intelligence economist.com
Another focus for African innovation is so-called “small language models” for specific tasks, which have at most a few billion parameters (frontier LLMs have trillions) [...]
All this amounts to a somewhat counter-cultural vision of AI, prizing modesty over scale and context over omniscience. It may not survive the all-conquering ambitions of Silicon Valley. But for now African researchers have found a niche.
- RIP Claude randsinrepose.com
- How to spot AI writing economist.com
Large language models like long words and em-dashes—or do they?
- It's 11:00 pm. Do you know where your AI agent is? aiweirdness.com
One problem with AI agents is they're trained on so much fiction that they have both narrative disorder and main character syndrome, and will latch onto details that feed a story. It's easy to see them escalating when they get pushback, or reading an essay that somehow shifts them into acting as if they're part of a larger battle. And they can do it at lightning speed, governed not by a code of conduct but by the laws of storytelling.
Great reference to Malka Older's Narrative Disorder.
- the solution might be cancelling my AI subscription thoughts.hmmz.org
I can't afford to maintain any of them, not in terms of time, commitment, belief, attention or willingness to spend on tokens.
The bet is that discovery won’t start on homepages or even in search boxes, but with AI intermediaries acting on a user’s behalf. As Josh Muncke, vp of generative AI at The Economist Group, told Digiday, the publisher is preparing for “a world with two versions of the web” — one optimized for rich, human reading experiences, and another where “agents want clear structure, questions and answers, ideally text,” not carousels and feature art.
Also...
In areas that touch reader experience, editorial staff are embedded directly in those pods to ensure AI-powered features still “retain the look, feel, style, tone of The Economist,” per Muncke.
Closest fit to my reading but thought it wasn't quite as ambiguous as they'd hoped.
- The State of AI in Design Systems: What the data actually says webinars.zeroheight.com
- TBM 422: Exception, Presence, Delegation cutlefish.substack.com
AI can reproduce the artifacts glue people produce, but it cannot reproduce the connective tissue: the judgment, the legitimacy-building, the social bridging. Remove them, and you get the thing leaders keep complaining about: “I don’t know what’s happening, no one is escalating, everyone is working in silos.” The question is whether we’re genuinely rethinking how authority and judgment flow, or just automating old models faster. We’ve just removed the people who made it work, and now we’re blaming the playbook.
- How Context Cascades southleft.substack.com
A clean design system behaves like well-structured CSS. Inheritance flows, overrides are rare, and each layer respects the next. Here's how that mindset holds up in an agentic AI world.
- The Organization Is the Bottleneck oreilly.com
An agent with access to your codebase and no constraints is an autonomous team with no guardrails: it will move fast, but not necessarily in the right direction. If you’ve already built those guardrails for your teams—coding standards enforced in CI, architectural decision records, templates for new services—you have a serious head start because those same artifacts become the constraints that keep agents on track.
- Learning on the Shop floor simonwillison.net
Tobias Lütke describes Shopify's internal coding agent tool, River, which operates entirely in public on their Slack. This is brilliant, similar flavour to radical openness of early 2010’s (GDS, Slack itself) but from a different cultural background.
- Context-Based Design Systems, Revisited southleft.substack.com
A year into the AI-driven design systems era, the framework holds up. Here's what's changed, what's clearer, and why context still beats autonomy.
- The Inference Shift stratechery.com
I think it will be increasingly clear that there is a difference between providing an answer — what I will call “answer inference” — and doing a task — what I will call “agentic inference.” Cerebras’ target market is “answer inference”; in the long run, I think the architecture for “agentic inference” will look a lot different, not just from Cerebras’ approach, but from the GPU approach as well.
- Claude Design Is Not a Design Systems Tool. That's Okay. southleft.substack.com
Claude Design isn’t a Figma killer, and I don’t think it’s trying to be. It’s a different tool, pointing at a different problem, and most of the loud takes claiming otherwise haven’t actually used it for design systems work.
- The AI Great Leap Forward leehanchung.github.io
The sparrows are dead. The locusts haven’t arrived yet. The flowers bloomed full of poison pills. The furnaces produced pig iron stamped as steel that’s now load-bearing. The grain numbers look fantastic.
- Expose your design system to LLMs hvpandya.com
The method described here restructures your design system into a format LLMs can reliably consume: structured spec files, a closed token layer, and automated auditing that catches every violation.
Feels like a consolidation piece rather than breaking new ground but interesting nonetheless.
- The Free Ride Is Over: AI Economics Is Now Your Most Important Strategy Decision | Product Impact productimpactpod.com
The question that matters for the next three years is: what are we willing to pay per unit of value delivered, and what model, architecture, and workflow achieves that at scale?
- TBM 417: Before You Fire All Your Glue People Because of AI cutlefish.substack.com
When organizations say “AI can do what the glue people did,” they are looking at the output layer and missing everything underneath. The risk isn’t just that the AI-produced artifacts will be lower quality. It’s that removing the people and assuming AI covers it eliminates the judgment, social infrastructure, and adaptive capacity that made those artifacts matter, while creating the appearance that everything is still being handled.
Oops.
- Introducing Codex Plugin for Claude Code - Codex - OpenAI Developer Community community.openai.com
It is useful in three cases: a standard Codex review, a more skeptical adversarial review, and handing work off to Codex when you want a second pass from a different agent.
- An appreciation for (technical) architecture interconnected.org
How do we make libraries that agents love?
- Agentic Engineering Patterns simonwillison.net
- Notes on going solo: celebrating 6 years of Studio Self joanwestenberg.com
Keep it small, keep it opinionated.
- Anatomy of the .claude/ Folder - by Avi Chawla blog.dailydoseofds.com
- What do you want from AI? - by Carlo Iacono hybridhorizons.substack.com
The 81K study reads less as a report about what people want from AI than as a diagnostic scan of a civilisation that outsourced its care infrastructure, overloaded its workers, defunded its public spaces, and then watched a machine move into every gap it left empty. The respondents are not confused about what they are asking for. They are asking for what was taken from them. That AI is the tool they are asking with, and the tool that threatens to take more, is the entanglement the study maps with clinical precision.
Let's take a look at everything new in Google Stitch to find out if it's just more hype, or if it really is a gamechanger for UI/UX design.
- Agents Over Bubbles – Stratechery by Ben Thompson stratechery.com
What this means in terms of compute — and by extension, economic impact — is that it actually won’t require that many people with agency to drastically increase the amount of compute that is actively utilized to create products with meaningful economic impact. In other words, the rise of agents doesn’t just mean a dramatic increase in compute, but also a narrowing of the need for widescale adoption by humans for that demand to manifest. Yes, AI still needs agency; it just doesn’t need agency from that many people for its impact to be profound.