Stanford researchers used a genome model to generate bacteriophage designs. They screened thousands, synthesized nearly 300, and found 16 that worked in the lab. A proof of concept, not a treatment. thefastnow.com/signals/sixte…

Caption controls should be easier to find on covered devices and services after a federal rule reached its compliance date. The practical test: how many clicks does it take on your screen? thefastnow.com/signals/how-m…

A four-week study found personal turns in prompted and unmodified ChatGPT-4o conversations. Ask who introduced the turn, what gets retained, what controls exist, and where a person should step in. thefastnow.com/signals/who-s…

The first #Android #app I built is installed on my phone. That sentence still feels strange. I do not know how to code, but I had an idea for software that could support a remote-viewing practice I have followed for years. The prototype let me start a session, record what happened, and draw directly on the screen. When I opened it and saw my own process working on my phone, I thought: I might have an actual product here. Something I can keep developing. Then came the awkward question. Did I build it? The best answer I have is: I built it with my AI. I brought the idea, my completed session artifacts, reference videos, knowledge of the protocols, and the judgment to recognize when the first direction was wrong. The AI handled the technical implementation and helped refine the prototype. Without those practitioner artifacts, it could have built an app. It would have missed the practice I was trying to support. That experience changed what I thought I was capable of creating. AI lowered a technical barrier that had kept ideas like this in my head. But the useful design still depended on experience the AI could not supply for me. For the builders and practitioners here: what did you have to show the AI before it could build something that fit your actual work?

An #AI provenance mark can show that material may have passed through a model. It cannot settle authorship, accuracy, or responsibility, and the trail may break during editing or resharing. The Label Is a Clue: hypernovelty.institute/artic…

An #AI provenance mark can show that material may have passed through a model. It cannot settle authorship, accuracy, or responsibility, and the trail may break during editing or resharing. The Label Is a Clue: hypernovelty.institute/artic…

#California's DROP portal lets eligible residents send one deletion request to 600+ registered data brokers. Processing can take up to 90 days, with exemptions and matching limits. What the new privacy exit does: thefastnow.com/signals/your-…

I almost shipped a book with a broken cover. Everything was lined up. Approved. Ready for the final #KDP upload. Then I looked at the cover one more time and realized it had reverted to an earlier version. Multiple rounds of edits appeared to be stacked together, and it looked terrible. I stopped the upload. The #AI searched the logs, found the approved cover, and fixed it. We missed the window for that day, so the release had to wait. That context failure was frustrating in the way AI failures often are. You explain something, move forward, and somehow end up explaining it again. Goldfish energy. But telling the AI to recover the right cover was still much faster than redesigning it myself. I got the speed and the mistake in the same workflow. Welcome to this place: amazon.com/dp/B0HDJMW6KQ This is why I keep a human reviewer at the points where digital work becomes a physical product, a public page, or a real expense. Those are the moments when someone needs the authority to stop the process and look at what is actually there. The AI handled the recovery. My eyes kept the bad cover from going out. Where does your workflow become consequential, and who has the authority to stop it?

I'm looking for five people to test Signal Atlas, an early Windows forecast journal. Crew Zero includes two feedback rounds. Completed testers receive bounded Founding-tier access for the life of the product. Windows 10/11, 18+. Apply: docs.google.com/forms/d/e/1F…

Great research paper on optimizing harnesses. (bookmark it) There is a lot of alpha in building a harness. And you don't need much to keep them optimized. This paper argues you can do this effectively using the harness own executions. The harness is the external control layer that turns a base LLM into an executable agent. Automatic improvement methods optimize a narrow part of it, usually prompts or pipelines, and deployed agents then reuse a single global harness for every case. MemoHarness decomposes the harness along the temporal flow of inference into six editable control surfaces (context, tool, generation, orchestration, memory, output) and turns improvement into structured editing over those dimensions. It documents per-case diagnoses plus distilled global patterns about what works and how dimensions interact, then adapts to each new case by retrieving similar past cases. No compute is waisted on test-time labels, feedback, gradient updates, or extra search. On the shell-agent benchmark it reaches 0.806 against 0.722 for the strongest fixed-harness baseline, at lower per-task dollar cost than the strongest commercial baselines compared. Paper: arxiv.org/abs/2607.14159 Learn to build effective AI agents in our academy: academy.dair.ai/

Using this as inspiration to expand my one man operation.

I built Polsia into a $250M company in under 3 months. Solo + AI. Zero employees. Everyone asks me how I did it. Introducing aisloP, a docu-series on how I build Polsia. Episode 1: The Launch. How I orchestrated the biggest Twitter launch of 2026.

Same pattern every time

Every revolution follows the same pattern. First they ignore you. Then they laugh at you. Then they fight you. Then you win.

AI for the Age of Too Much is live on Amazon. It’s a short practical manual for staying in control while delegating to AI — built for people who feel buried by tools, prompts, and automation advice. amazon.com/dp/B0H3738LW8

Build your one person company with support.

#AI is starting to move out of the “tool” category. The better question is no longer only: what can it generate? It is: who owns the consequence after the system acts? New #Hypernovelty Field Note: When AI Stops Being a Tool Story open.substack.com/pub/hypern…

Check out my latest article: The Lightning Network Is Quietly Becoming the Rails for AI, Stablecoins, and Global Payments linkedin.com/pulse/lightning… via @LinkedIn

You are not your body. No single atom in your body is the same as when you were born, yet you are still you. Because you are the consciousness that uses your body as an avatar to navigate in the 3D world...just like how you use an avatar to play a video game. All suffering arises when we forget that we are not the avatar, that we are outside of the game. And science is even proving all of this ancient spiritual wisdom true. In 2022, the nobel prize was given to someone proving the universe is not locally real -- meaning that what you look at isnt actually there, it's just rendered via your consciousness every time you do. A table is only a table because your consciousness renders it so.... And the SAME goes for your body. Wild shit. This is how out of body experiences and beautiful near death experiences can happen...because they release us from this false identity with the body. Use your body as it's meant to be -- to play in this world and evolve into your highest self, to discover who you really are. But if you get overly obsessed with it, try to make it live forever, you will end up miserable Realize who you really are, beyond the finite body, and be free. MOOO