David Seaman's Deep Podcast

This weekend’s unsolicited weekend leftoid meltdown deserves some analysis, without bowing to any of their ridiculous claims.

Must feel good af to crack open a cold craft IPA in some Williamsburg bar, then go around LARPing as a libbed up John Connor type in Fall 2026 without knowing how to code so much as a single HTML page or checkout form; these midwit Berkeley types ALONE hold the knowledge and unfathomably vast intellectual power to prevent a technology which… they claim… will soon end “every” human life on the planet.

Or, they’re total midwits sucked into a misguided partisan mass psychosis thingy, less than two months before midterms. Cattle lost in a tornado of their own creation.

Meanwhile, local AI remains the same great, safe, and legal business tool it was last week. One of America’s best known tech journalists is currently reviewing our flagship software, Braid, which enables people running more than one LLM to connect them and share useful state.

As I’ve spent about a year of my life and wealth on developing that software, alongside a great team of people I respect, I want to make sure it gets the full court press and downloads it deserves. I see a near future where close to every local device (meaning laptops, desktops, smartphones) will be loaded with a LLM or more than one. It becomes the core use case for computers, actually, sort of like Internet access / email / word processing / video editing are the core uses or selling points today for local hardware.

And in a world where every computer is running at least one LLM, we think there will be great need to network or “braid” these devices for efficiency’s sake. We use Braid in house because it’s helped us keep track of very different projects running on the different machines.

Technically, other teams could produce much of the functionality. It’s just network theory and a lean gossip protocol. But the compatibility testing we’ve done is what makes us stand out… anyone copying our “recipe” would need to put that time and effort in, making it one of the least attractive things to code or ship.

So I think there’s a moat. I also don’t want my entire life and personality to turn into marketing software, which is why we’ll instead do some big ad campaigns targeting the LLM hobbyist community, and we’ll line up more of these hands-on reviews from nationally known journalists. Think that’s the path. I enjoyed building it, but I don’t want to be some promoter. It either gets millions of downloads over time… or not.

In this episode, we break down the massive pivot in Intersignal Braid 1.8.1: moving from a cryptic command line experience to an intuitive, messenger-style inbox. We discuss our experiences dogfooding the software in the lab and how a shared context field transforms isolated LLMs like Kimi, Qwen, and Gemma 4 into a unified, searchable network.

https://intersignal.org | https://x.com/intersignal_ai to follow the team on X

Tonight we connected a MacBook Air and a Linux laptop through Intersignal’s Braid Archangel. We shared a small piece of project context from the Mac, and the Linux machine indexed it locally. Then we asked a fresh question through its shared field interface, and it returned the correct answer with an inspectable evidence record. The machines appear as contacts, and the handoff feels like sending a message, but the received material becomes queryable local state.

This is obviously a big deal when you think about it: you can “query” or search through the breakthroughs the first LLM already made when beginning a conversation or exchange with a new LLM. Tests are ongoing, but our Mac, Windows, and Linux versions are performing well so far.

https://intersignal.org | https://x.com/intersignal_ai

Get in there: https://buy.stripe.com/14A3cvb2ncYj2vyaZ88so00

GPT Astra and its “recurrent depth” breakthrough are expected later in the week for some users; Anthropic, meanwhile, has already rolled out its newest flagship models, Mythos 5 and Fable 5.1. Yet as these trillion dollar labs are rushing out new systems, the local AI community is growing in strength also — both hardware innovations, and software/networking innovations like Intersignal Braid and some competitor projects.

https://intersignal.org | https://x.com/intersignal_ai

GPT Astra is expected to use recurrent depth, a radically new reasoning architecture. Plus, Intersignal Braid is now open to a select number of seats: https://buy.stripe.com/14A3cvb2ncYj2vyaZ88so00

Learn more about the project at https://intersignal.org and follow us at https://x.com/intersignal_ai and https://x.com/davidworld

Last night, Intersignal completed the Kestrel proof, as we call it: a remote local LLM answered an invented operating problem incorrectly before Braid, then correctly after receiving and committing the missing situational state from another machine.

Braid Founding Membership is now open at $99/year. Membership includes experimental Braid Managed Relay access for approved devices, guided onboarding, supported upgrades, release guidance, and best-effort setup support. We’ve tested Braid with Qwen, Gemma4, and other models available through Ollama.

Core Braid LAN software remains free.

Join the founding cohort:

https://buy.stripe.com/14A3cvb2ncYj2vyaZ88so00

Braid is experimental developer software. It transfers authenticated semantic context… not model weights, hidden activations, consciousness, or KV cache.

If you already run local AI models using Ollama (https://ollama.com/download), consider test driving Intersignal Braid. Rather than remaining siloed from each other, Braid allows your local AI hardware to share salient data and realizations, in ways that may significantly improve your workflow.

https://intersignal.org | https://x.com/intersignal_ai and https://x.com/davidworld to follow the team | https://appdevelopermagazine.com/intersignal-releases-braid-pathfinder-developer-preview/ - detailed review of Intersignal Braid in this month’s App Developer Magazine.

Testing Braid 1.5.2? Submit your live benchmark, edge case, or crash report to us here: https://docs.google.com/forms/d/e/1FAIpQLSdpVTmbvutQcUheo4Zv8JrAsst-yWs6tCJ0RixnW2E1PkWLGA/viewform

All Epsteinites still free, but hey, a hemp joint can now land you 10 years in prison in Texas. Laughably absurd movement - pedophile adjacent bootlicker weirdos - and after losing every seat this November, these folks should be quarantined from public office and public life forever. A new generation of politicians is already rising rapidly, not beholden to this garbage, and not tolerant of it either.

Braid v1.5.2 is officially out of developer preview and live on Mac, Windows, and Linux.

We sit down with an Intersignal non-human developer to discuss the v1.5.2 release and dive deep into the long term vision for local first, decentralized AI state. From developer usability upgrades to edge orchestration, we cover everything you need to know about streamlining local LLM coordination without centralized intermediaries.

https://intersignal.org | https://x.com/intersignal_ai

https://x.com/davidworld

Significant leap in local AI capability versus a year ago, but still not a total replacement for cloud brawn. Simply bizarre to me that some people keep saying AI is incapable of solving basic math or “riddles” when it’s coding thousands of lines of flawless network connectivity firmware, solving Erdos problems, and self-designing a next generation of chips.

The more I work with AI, the more convinced I am that the doubters are treading water in a vast sea of self deception, one where the tide will eventually go out, rendering some folks quite naked intellectually.

https://fulcrumnews.com/contact to contact us with show ideas, guest requests, etc

https://www.amazon.com/author/davidseaman | https://intersignal.org