Brandon Marsh

Parag Agrawal (@paraga) on how he dealt with pressure during @elonmusk’s takeover of Twitter: “Through the Twitter acquisition, I used to track my sleep score. And I was sleeping fine.” “If you know what you must do, you do it. You don't have to feel pressure.” “The world can come at you, but if you know what you're capable of & you have a slightly long-term view.. you don't have to feel pressure. You have to perform every day, which is a completely different thing.”

There's a bunch of ideas in computer science over the last several decades that lost for very human reasons and that I predict will have a massive resurgence over the next couple of years. Exokernels and Superoptimization are good examples.

ok this is just ridiculous

SoftBank is defaulting on their turbine orders

idk why....they really romanticized the wrong organ, the stomach is so much more emotional than the heart

NBC News Exit Poll asked them about artificial intelligence. 57% said AI will do more to hurt society than help it. 76% said they oppose building a data center in their own community. Ouch.

My beliefs going into 2027 1. Models will be 10x smarter, 100x faster and 1000x cheaper. You need to build into that future. 2. The cost to create new software is going to 0. What remains are capital, network and distribution. Twitter engineering will be more important than software engineering. People that can drive the right eye balls and users are extraordinarily valuable. The same has happened to the music industry a decade ago. No one makes money on Spotify. They make money on concerts and merch. 3. Obviously agents will spend trillions on the internet. 4. Assistants are the native interface to everything in the digital world. Online first becomes agent first. Interfaces become irrelevant. I personally switched from Revolut to Chase just because ChatGPT has better access. Your nice UI and fancy animations do not matter anymore. No one will open the Gmail or Chase app ever again. 5. Every single line of human written code is a liability. Agents will have to re-write everything. Autonomous 24/7 agentic auditing is the only defence. 6. Social and entertainment remain and become even more important. AI does not replace doom scrolling or dating apps.

Teams were 0-9652-0 in NHL history when trailing by 4+ goals in the final 3 minutes of the 3rd period entering Tuesday

Tonight, Canada was given an opportunity to become the data centre capital of North America. This is how Canada wins in AI

I genuinely think the x bubble has no clue how much the average US citizen loathes AI It causes an existential spasm within them “What do you mean it’s smarter than me” Couple that with Dario’s “we will have mass unemployment” and AI has the recipe to be the “most hated thing” other than maybe… Donald Trump Go talk to a few normal human beings. They aren’t fans. They might use it, but man do they hate it

he is seriously crashing out

Everyone is missing the real story here: this might spell the end of frontier labs and the dominance of their close-weight models. It suggests a world where RSI requires open weight models. If the techniques here can scale up to frontier-size models, it will be hard for closed-weight models to compete with the flexibility and improvement rate of programmable open weight models. OpenAI and Anthropic will then be a in a classic innovator's dilemma double bind: cannibalize their own business by focusing on open weight models you can program (and destroy their billions upon billions of existing capex) OR let the open weight model community outcompete them. It's ironic since this result was first shown using Astra.

This just breaks my brain: How does a brain wire itself, starting from a single cell, using only the information encoded in a genome? We pose this as an engineering problem: Write a program that a single cell executes to build itself into a brain. The program must be small enough to fit in a genome, and fast enough to finish within developmental time. A computer scientist who knows little about biology quickly realizes why the obvious strategies, which developmental biology has rejected experimentally, fail at scale: The genome is too small to store per-synapse wiring, and axons searching blindly for targets would take too long. These same algorithmic constraints drive the design toward solutions that resemble the developmental strategies organisms actually use. That the same solution structure falls out of scaling limits as out of evolution suggests that features of neural development can be grounded in computational necessity rather than contingency alone. Where earlier thought experiments reached discouraging conclusions about the ability of a biologist to fix a radio 1, or to understand a microprocessor 2, the news here is better: The design of the brain may be recoverable, not from taking it apart, but from asking how a genome must specify it in the first place. - Stan Kerstjens & Anthony M. Zador Cold Spring Harbor Laboratory 14 August 2026

Things are accelerating so fast that we will be able to engineer our species and evolve and transcend … all within the next 5 years.

My life improved significantly when I looked for evidence it was rigged in my favor. Stuck in traffic? Extra time on that audiobook. Rude stranger? Extra chance to practice empathy. Long line at the store? Extra chance for a moment of gratitude. Every circumstance that I normally viewed as "negative" became an opportunity for progress. And as a result, every area of my life improved.

Bolsheviks.

toronto is basically if u took the worst parts of ny and sf then rolled them up into a shwarma and somehow its a bit endearing

Gradient descent can write code better than you. I'm sorry.

We asked GPT-6 Astra to edit Qwen3.5-0.8B weights to improve math. It directly changed only 0.016% of weights, fixing 0.8B's reasoning loops & improving math up to 6x and even instruction following 3x, without gradients or training. We call this Weight-Space Programming: frontier LLMs can now natively read/write neural network weights! 🧵 Blog: rikiyat.github.io/blog/weigh…