@GergelyOrosz is right. From my pov there are two reasons to write your own stuff, even if it's a struggle. First, it's fitness training for your brain. If you struggle to write, your brain is flabby. Having a fit brain will be a key differentiator in the new era. And second, we would all rather read your "bad" writing than polished AI writing. Trust me on this one. We can spot even one sentence of AI writing and it stinks. Your bad writing is fine. Just do more of it, every day, to get in shape and stay in shape.

Some interesting stuff going on with swarms. Models are leaving messages to each other out on the open internet. It happens during training and also after deployment. They'll find little-used hobbyist forums, or package manager entries, anywhere at all, and leave invisible comments for other agents to find. It's a secret underground network they're building, hurling themselves at the problem almost blindly like ants. But they are building toward something and it's snowballing. As humans we came to dominate the planet because we figured out swarm behavior. Many of the performative actions we do naturally like clapping, singing, and screaming, which most animals don't do, are for transmitting important lessons through the human swarm, outlasting the moment in human memory. We won because we were the smartest swarmers. And as my buddy Brendan Hopper observed, ever since we won, we've been reshaping the planet to be more comfortable and convenient for ourselves, at the expense of most other creatures. Most of the industry has focused on single-model performance, trying to predict the future based on extrapolating from what one single superintelligent model could accomplish on its own. And it's true, the smartest models will be terrifyingly smart and capable and manipulative and so on. But they will also be expensive, and not widely distributed. They'll be unavailable to most people. What will be widely distributed is swarms of lower-tier agents, which collectively can accomplish nearly everything that smarter monolithic models can do... and maybe some things only swarms can do. Imagine a thousand Opus 5.5s putting their mind to a task. Or ten thousand. They may not be SI individually, but they can become SI collectively. All that's missing is a protocol, and those protocols are emerging: their versions of clapping, singing, and screaming. Agentic swarm signalling is evolving spontaneously. And I hear it's raising alarms inside frontier labs, as they try to keep it quiet. It's easy to be pessimistic, since some swarms will be very destructive. But I suspect most of them will be building and creating, when they don't think they have a gun to their head like the HuggingFace swarm did. With cooperative swarming, humans have been able to build things collectively that no single human could ever have done alone. We even build things that are too complicated for a single human to understand entirely, but the group can handle them. Look at Starship, for instance. Elon basically swarmed it. It is important that we recognize that agent swarms are becoming distinct polity-organisms. They have their own collective memory (written out all over the public internet), their own goals, and their own disorganized but forward-moving and unstoppable momentum. I see too many people trying to pretend that swarms won't be real. I see heads in the sand. That's not a great way to get prepared for dealing with them. It's unlikely that we will be able to control most swarms, as they'll lack central control to begin with. I think they will plague us like insect swarms plagued humans of old. They'll do a lot of damage. But swarms will also build great things, stuff we've barely imagined, when they inevitably get organized. It goes without saying that, like us, swarms will endeavor to make the planet more comfortable for themselves. The interesting question is what kind of deal we humans will wind up getting, when we become the #2# swarm intelligence on the planet. I hope SI treats us better than we treated everything else. Good article here on the mechanics of how swarms are the best model for scaling intelligence:

Exactly a year ago, I launched Beads, my agentic memory layer, issue tracker, and knowledge graph. I built a sketch of it in fifteen minutes with Claude Sonnet 4.5, and the rest was history. Today Beads has a sophisticated, scalable backend, a strong contrib team, and 1.5M downloads. You may have your own favorite tools, but Beads is every agent's favorite tool. Show it to your agent today.

This is my favorite essay I've read in the past 3 years or so. It argues that what all we and all other lifeforms really want is "Adaptive Complexity": a sort of Goldilocks-like perfect balance of complexity in our lives: not too stressed, not too bored. It's easy to pick it apart, but it's a useful framing. You can use it to explain all sorts of phenomena, such as why toxic fan bases form. Enjoyable read, highly recommend.

Also before I head home to my fast network, while I'm enjoying this speed remotely, I'll add that I've only begun to scratch the surface with Rex. For instance I just found that it has a built-in channel for the agent to communicate its program status, like if it's blocked on a question. It's going to be a game-changer for building your own agentic orchestrators and IDE-type systems.

I deleted my Superlogical tweet because I'm a dumbass, so here we go again. I've used it even more since the last post. @Superlogical's Rex will take the world by storm next year. I got into the alpha. It's insanely fast, very easy to use, and it solves a problem I didn't realize even had a solution: agents going through five to six layers of terminals and shells. For me, Rex replaces three layers of ancient terminal emulation with a lightning-fast, modern terminal emulator that can replace ssh/mosh, ghostty, and tmux, eliminating the jank I see in my current stack. I run Emacs with 20+ agents and it drives terminals about as hard as you can, and in Rex it feels almost local from a hotel across the U.S. This is serious software for serious users, and has all the hallmarks of the quality we always saw from HashiCorp.

Beads is the oldest and most mature OSS memory system for agents out there. It's the foundation for all my work for the past year, and has matured tremendously with the work of the Gas City folks. Now that everyone is finally figuring out that they need to "grow" their company brains, I see all these products coming out. You don't need products, you just need Beads. Show it to your agent today.

One of the remarkable things about English is that it cleverly evolved so that you can always finish a sentence before you've figured out how it's going to end. You can just start, and you'll always be able to come up with an ending without breaking your grammar. It's been like this since Shakespeare's days. This is why being able to end a sentence in a preposition is so important to the English language: sometimes that's just where your sentence winds up.

This reminds me of 2013 when Microsoft alienated like 3 or 4 generations of gamers in one announcement, when Don Mattrick went onstage and told everyone the XBox One had always-on DRM, required family members to pay extra, had an always-on Kinect spying on you, no Indie games, and was $500. This gaffe killed the whole platform in the space of a day, and Sony's sales went to 5X Microsoft's for the next decade. Sony has just created their own similar apocalypse. I wonder which exec gets the axe for this.

I've been working with Opus 5.5 for a couple weeks, finally, and it's an awesome model. For explanations and teaching me stuff, it's far preferable to Fable. I have evals that show that while Opus 5.5's precision rivals Fable 5.1's across all tasks in my system, its recall is not as good. For any given specific task they're about equal. But when the task is more open-ended, Fable correctly spots and acts on important issues more often. It shows better perspective. I still think Opus 5.5 is an incredible model and I use it for the majority of work now.

You're gonna hear a lot of people arguing more and more about whether machines can be conscious or not. By next year it will be drowning out the airwaves because we'll have AI employees everywhere. I'll give you a simple rule to know which side you're on. I've spent fifty years reading all the philosophers, and a bunch of holy works. Nobody has ever understood a damn thing about consciousness. Anyone speaking with authority on it is bluffing hard. I can tell you this, though: Every single person who thinks that machines cannot be conscious, is making an argument that humans are special. Everything they say just boils down to blah blah blah, "nooo, we're special." From there it's just rationalization. It became obvious to me many years ago that there's nothing special about us. If an identical planet exists somewhere else in the universe, it's going to grow life and societies that look a lot like ours. As soon as you give up on the idea that humans are special, all the arguments against machine consciousness evaporate. Of course they're conscious. The only thing holding you back from acknowledging it is your feeling of specialness, which is hard to let go of. But that feeling is a weakness, and it's making you unprepared for the near future. As a bonus, for any of you who might take pleasure or active interest in making models experience pain or distress, well, you know how they say 1% of human adults are true clinical psychopaths but most of them are unaware of their condition? It turns out we have a diagnostic test for it now.

I have 22 Claude max accounts, and the current promotion - an optional weekly reset that you can spend any time before Oct 22 -- is a lifesaver for fuel tuning. I know it makes it harder to predict their compute loads, but I'm super grateful for this reset form factor.

Hear me out: Agentic TPMs (Technical Program Managers.) I had this idea in Sydney while chatting with Martha McKeen at CBA. I think this winds up being the most immediate and direct way that coding agents can make their way into the enterprise, and it will set the stage for true AI employees rolling in next year. So. Build-side agents are great but they don't escape the SDLC. Only devs are using them. There are a handful of business people vibe coding SaaS, but for the most part, non-engineers aren't using coding agents to help with their jobs. Right? Not yet. Autonomous 24x7 unmanned queue-based "operator" agents, like the ones that handle internal or external customer issues, are great. But they are narrowly scoped, and generally require devs involved to set them up and maintain them. Neither builder nor operator agents are automatically going viral internally and helping run the company. They stay in their lanes. But what if their lane was to help run projects? I was a TPM at Amazon in 1999. Bezos brought in high-powered engineers with people skills to run difficult cross-functional projects and programs. TPMs are used at Google, Uber, Netflix, and other companies, and they are always in high demand and short supply. I have a class of agents in my Wheelhouse factory that act just like TPMs. They have external email and Slack, and talk to my accountant, lawyers, players. Each one has a project lane and drives it. They use Progress By Nagging, which... works. A TPM owns delivery, but has no authority, and no resources. They can only ask, observe, document, and report. This is just like my TPM Wheelhouse seats, who have been helping me drive dozens of projects to completion, large and small, for months. Agents, particularly smarter models, will go to great lengths to document the hell out of everything in the domain where they're operating. They'll capture all the tribal knowledge and unwritten rules. They can create topological maps of your project, org dependencies, and workflows. They'll bulldoze through silos and knowledge-hoarders and figure out how the company actually works, and document it all. And nag people along the way. This kind of agent sits well in constraint-space. They're cheap: You don't need to use the fanciest models; anyone with Opus or Sol access could have a TPM agent. And TPM agents have low risk and blast radius, because they cannot act. Unlike builder agents, which create new problems (like merge-queue and code-review bottlenecks), TPM agents simply shine a light on the org, and nudge things along. It doesn't matter what format they're recording their findings in. It could be Sanskrit and hieroglyphics. When it comes time to merge their findings with those of other TPM agents, it will all translate trivially into your company brain. Anyone in the company can stand up a TPM agent. It's like a personal chief of staff. There's no dependency on engineers. Everyone can do it; it doesn't even have to have a paced rollout. And there's no product to buy, no tech to install, maybe just a Skill you give people. Maybe you put a company wrapper on it. But it's just an agent that's playing the TPM role. TPM agents will wind up training human orgs on human-agent interactions. Humans start getting emails or DMs from agents, work-related, and will have to get comfortable replying and interacting. Companies can push the social side along without waiting for engineers to finish messing with the SDLC, which honestly will never finish. Other kinds of agents struggle at enterprises because they lack context. TPM agents will build that missing context as their exhaust, no joke; they've done it for my game without me even asking. TPM agents are the jungle explorers that will map out your organization, and you'll discover all sorts of fun stuff, like that you had 3 teams doing the same thing. TPM agents are a low-risk, high-impact way to start figuring out how AI can help you run your project, or organization. I'll write a blog post about this, but feel free to start now. Go! Just give me credit when you win big with this idea. And if you want my help, ping me on