Saw this posted on LinkedIn. I guess I'd probably agree.

"I've donated millions of dollars to Democrats, I've never been to the Oval." Your political donations should not get you - or anyone - a trip to the Oval. They should not get you an audience with the president. They should not get you calls from the White House to ask your

I don’t give a hoot about all the low-IQ techbro grifters who have lost all the credibility over the past few months. But it’s particularly disappointing to see one particular tech “guru”, whom I used to look up to and respect, embrace and even double down on all the idiocy.

For years, my favorite trait in developers was curiosity. Curios developers learn fast and write more reliable code because they follow all the small "mmm.. why does this happen" moments before users see them. But I have a new one now. Diligence. Developers that maintain

Press release: “shaping the future of AI interoperability and seamless agent collaboration”. Reality: stuffing JSON as strings into LLM prompts.

I think AI is going to usher in a gold age of infra, not obviate it. It's just so clear that good CS fundamentals result in better AI built systems. Vibe coding works better with type safety, languages where syntax maps closely to semantics, referential transparency, tight

It's striking how much Trumpists hate what's made the United States great: immigration, trade, medical and scientific research, the rule of law, an enlightened patriotism, a willingness and ability to correct past injustices. At the end of the day, Trumpism is Third Worldism.

The lesson of the 2020s, like the 1930s, is how much damage bad leaders can do.

In the 3 weeks since the tariffs took effect, ocean container bookings from China to the United States are down over 60% industry wide. 🧵

Let me explain: When you merge all commits from a PR you get all the history of every change, often including multiple iterations on the same code blocks - none of which were ever “in production” in isolation. If other people work on the same part of the codebase, they have to

You know what the single most profound, far-reaching effect of Russia’s war against Ukraine is? Ukraine has finally learned to truly respect itself. We are no longer some godforsaken "Russia's backyard", where corrupt mafia clans don’t dare to make a sound without the Kremlin’s

I have worked with pretty much every major AI model/provider. They all have strengths, but man do they have some weird issues too. I've been keeping a list. Decided to share because why not? Here's everything I hate about Gemini, Claude, ChatGPT and Mistral.

$20K/month is where your AWS bill changes the most. •$6K on RDS IOPS. •$4K on transit charges between AZs you forgot •$3K on NAT gateways passing idle traffic •$2K on S3 PUTs from the microservice the principal engineer wrote •$5K on CloudWatch logs

I don't understand humanoid robots. You can give it a shovel, and it can dig a ditch. Excellent >100 years ago, but today we have excavators. The same goes for heavy mass farming equipment, robot vacuums, and factory automation (video below). Why are people excited?

The Illusion of Thinking in LLMs Apple researchers discuss the strengths and limitations of reasoning models. Apparently, reasoning models "collapse" beyond certain task complexities. Lots of important insights on this one. (bookmark it!) Here are my notes:

𝗔𝘀𝘁𝗼𝗻𝗶𝘀𝗵𝗶𝗻𝗴𝗹𝘆 𝗳𝗲𝘄 𝗽𝗲𝗼𝗽𝗹𝗲 𝘀𝗲𝗲𝗺𝘀 𝘁𝗼 𝗵𝗮𝘃𝗲 𝗮𝗽𝗽𝗿𝗲𝗰𝗶𝗮𝘁𝗲𝗱 𝘁𝗵𝗲 𝘁𝗿𝘂𝗲 𝘀𝗶𝗴𝗻𝗶𝗳𝗶𝗰𝗮𝗻𝗰𝗲 𝗼𝗳 𝘁𝗵𝗲 𝗻𝗲𝘄 𝗔𝗽𝗽𝗹𝗲 𝗿𝗲𝗮𝘀𝗼𝗻𝗶𝗻𝗴 𝗽𝗮𝗽𝗲𝗿. I will break it down for you, in four sentences: • It is not just that these

Another excellent post by prof Misra on the hard information theoretic limits of AI, including: - limits to long term agentic behavior - limits to long term reasoning - limits to self improvement

Reward models (RMs) are the moral compass of LLMs – but no one has x-rayed them at scale. We just ran the first exhaustive analysis of 10 leading RMs, and the results were...eye-opening. Wild disagreement, base-model imprint, identity-term bias, mere-exposure quirks & more: 🧵

Every founder you admire has quietly survived moments that would make most people quit. There’s no playbook for sleepless nights, hiring mistakes, or explaining to your family why you missed another weekend. It looks like freedom, but most days you answer to your own ambition,

How many more model releases do we need for folks to realize we are not getting to magical superintelligence with what we got? How many times do you have to see a model benchmaxxing to realize Humanity's Last Exam is a freaking idiotic name and that answering questions on it