Memory Is Not Judgment The most interesting agent papers this week are all circling the same problem: an AI system can possess more information, more context, more compute, and more freedom to…

After California closed a previous regulatory loophole regarding autonomous vehicles and began allowing autonomous trucks to be tested on public … More

AI models are not behaviourally neutral infrastructure. Their answers reflect patterns of judgement, confidence, caution and communication - and those patterns can become material business risks.

Updates Below are some of the highlights in AI, robotics, and machine learning for this past week. Note: Because of the escalation in the velocity of AI updates, I now restrict these summary posts to…

This article shows how to build an Autonomous AI Agent with SqlFlow

"It was the best of times, it was the worst of times." This was written by a human. "The era was one beneficial to many people, but substantially detrimental to others." This was written by…

An approachable introduction to AI agents and why they’re likely to be the most impactful technology of our generation.

From The Lancet: USAID funding has significantly contributed to the reduction in adult and child mortality across low-income and middle-income countries over the past two decades. Our estimates show…

Anthropic’s best AI model struggles to attract users as cheaper tools thrive A few interesting numbers in this FT story gathered from "people with knowledge of the matter": Anthropic's "annualized…

A couple of months ago (maybe it was a year ago already? Who knows, time is weird in this day and age) I started to work on a little project, as I often do. The project’s name is Boris, which stands…

https://austinhenley.com/blog/aiadoption.html

Posted on August 23, 2026 by psu Earlier this year I spent a few hours transcribing all of my angry thoughts about the idiotic industrial hype complex around using the new “““AI””” (LLM) tools for…

Tom Blomfield’s five elements of a self-improving loop: sense, decide, act, verify, and learn.

Prior to Fable, it felt silly to waste too much time improving your coding harness or context strategies. A new model would arrive at the same price (or cheaper!) and paper over most of your…

TL;DR: Write the failing test first and ban deletions, or the AI deletes your test, reverts your fix, and calls it done. Common Mistake ❌ You ask the AI to fix a failing test, and it deletes the test

Today, congestion is partly constrained by human tolerance. At some point, a trip becomes slow enough that the marginal driver decides it is not worth making. But what happens when the marginal…

There’s some talk today about how agentic coders are balking at Anthropic’s pricing and adopting alternatives. I was reminded of a thought I had in the weeks following Fable’s release: the free lunch…

August 23, 2026 Anyway, what I really want to remark on here is the idea of using AI to brainstorm . In the past, I may have written this off as well. But actually, with the rise of the very good AI…

I use AI pretty much every day now, but I pair that speed with backups, human approval, and guard rails designed to keep mistakes from becoming disasters.

In November 1967 Italo Calvino gave a lecture called Cybernetics and Ghosts where he talked about writing machines and the future of literature: ...will we have a machine capable of replacing the…

An update on this post, from my inside sources: word on the math streets of SF is that the OpenAI team tried something like 500 problems to get their 10 solutions. I don’t know how much of an…

LLMs are prediction engines but that, on its own, cannot dismiss their intelligence.

Disclaimers The current discussion in Debian aroun the AI GR is very heated, and I won’t add to that, however, I am very confused about some of the viewpoints there. But, I had no idea how to even…

A series on how software architecture shapes AI-driven code degradation. An objection to the last post turned out to be right, and fixing it produced a better result than the one it demolished. Every…

I have finally integrated an AI agent into my daily work. It helps me to perform daily prioritisation and strategy brainstorm. I have been experimenting with AI for many months; it is only until…

A series on how software architecture shapes AI-driven code degradation. This post reports the strongest result against the position these posts have been arguing. It replicated in both runs, so here…

Concept from Ready Player One -- you enter a movie in VR and you have to play along as one of the characters in the movie, say the right lines, etc. Rick and Morty S8E7 ("Ricker Than Fiction")…

Robots break records in China, Anthropic continues super-growth in revenue, and Stripe buys OpenRouter.

A year ago I read Libraries of the Mind by William Marx, and some of it is still very much on my mind ever since. Particularly what William Marx called ‘dark matter’ literature, which…

I’ve been using Opus a lot at work lately for a project and have had serious issues with it reaching for bizarre analogies, obscure jargon, and strange phrases.

Google's Open Knowledge Format is quite a bombshell standard. No RAG, no nothing, just a wiki in a git repo is enough to make LLMs work! Some progressive disclosure by very basic dir/file naming…

The Library of Congress has crowdsourced more than a million transcriptions. A half-finished 1917 telegram at the National Book Festival shows why AI can accelerate archival work without replacing…

So, there is a great perspective de aprender com o Karpathy ainda. Vídeos que não assisti por completo e vídeos que acho que não comecei. And, important to consider these resources. A great…

At The Atlantic (gift link), Tyler Austin Harper writes about “Why the Closing of Harvard’s Writing Center Matters.” He suspects that Harvard’s decision will give other schools permission to close…

I prefer interacting with good AI customer service agents. I appreciate having the option to talk to a human, but I’ll probably take a subpar AI agent experience over sitting in a queue for 30…

There’s a lot about the current AI situation to be upset about. Data center buildup is insane and, certainly at a higher rate than we will need when the market settles. While they’re…

The automation of mathematics has been going on for more than a century. If AI can automate some or all parts of mathematical proof, that will be another important phase in this process, but not the…

We see variations of this question: What happens when robots can do anything physical better than humans and AI can do anything cognitive better than any humans? What will be left for us? The answer…

A decision can be fully explained yet impossible to correct when nobody has the authority and duty to own it.

If we look at using LLMs to write code and you have seen your fair share of similar technologies, there is an obvious reason why enterprise development seems suitable: semantic lightness to using…

I launched another blog yesterday: Little Muddy. For the next thirty days, GPT-5.6 Sol will choose a topic, research it if necessary, write a post, and publish it once a day without any intervention…

AI stocks will likely drag everybody else down very soon. As a student of history, you may already know this pattern has persisted throughout history. We are not ready for the type of AI being…

When the right action is unclear, progress comes from a simple loop: act, observe, learn—and change what you do next.

The AI bubble refers to the concept that the excitement about AI has led to excess investment. Many areas are contending with new data centers being built. The idea is that huge models, with…

In the final episode I talk about open weight model harnesses, or how to actually talk to your local AI. From llama-cpp chat to Pi, or Hermes

The message of coding agents is not automation, but the expansion of what organizations can express through software.

The concluding chapter of my SPIRIT series. 🙂 Let's start off with a recap: Substrate Self: The foundation. Team: Shifts the bar from whether the documentation is good to our foundation is…

Tech Industry & Culture Link to heading Dogshit In the late 80s, a Microsoft manager coined “eat their own dogfood” to push internal use of LAN Manager. The practice made sense when teams needed…

Dario Amodei may be right about AI and still be wrong about medicine. His argument is that intelligence is about to become abundant. Millions of AI scientists could read the literature, generate…

Most AI content is worthless because the person publishing it had nothing to say. The fix is not a better generator. It is starting from what you already know.