BREAKING NEWS: Anthropic's latest model will NOT help you if it thinks your ML research/ML engineering is interesting, and/or will secretly degrade its IQ so that the average engineer won't notice. We are already seeing Anthropic's latest model's moderation filters our GPU inference research and programming 😭

Random circumstances at birth shape beliefs that last a lifetime. Then we spend years reading books, hiring coaches, doing therapy, meditating, and diving into psychedelics ... all in pursuit of a proud, rare lifelong achievement: changing our minds, just once, about 1 thing

Awesome, SpaceX has begun exploring using Starlink to enable high-bandwidth around the Moon! Traditionally, deep space coms has relied on radio frequencies. This will enable gigabit connectivity anywhere on the Lunar Surface. Imagine watching livestreams. It’s gonna happen.

My experience with the last 10 PRs: Codex with GPT-5.5 implements plan and it passes right off the bat, both Claude Code Opus 4.7’s PR review and its own review (new context). Claude Code with Opus 4.7 passes neither its own review nor Codex GPT-5.5’s review.

Some good laws for working with AI. It's a good start. susam.net/inverse-laws-of-ro…

the guy who single-handedly raised Anthropic's engineering productivity by 70% just dropped a new podcast. Boris Cherny laid out the 9 core patterns he changed in his workflow to get there. the 3 most common traps: > bloated CLAUDE.md files that tax every single prompt before you've typed a word > re-reading old chat history on every message: by message 30 you're paying 30x the cost of message 1 > plugin hooks you installed 6 months ago and forgot, silently doubling your bill i'm confident 99% of you have hit at least one of these patterns and just shrugged it off. if you want to actually upgrade how you work with AI and Claude, watch this podcast to the end, then lock the knowledge in with my breakdown

AI is math, algos and compute. None is going to go back in the bag unless we fall into a dark age, as a civilization. Most likely we'll go forward. You can't stop math, algos and compute spread into the world. Stopping AI development in any part of the world is illogical. Nobody can stop that

Overheard in a SF coffee shop. "Man, you have a family and kids. Don't pivot into talent sourcing. That hill is filled with dead companies. You won't raise money, you'll struggle to find clients, KPIs don't work in your favor. Don't do it. Think about your family and your future"

Google wants us to Gemini, a little bit, just not too much. Money won't buy you quota anymore. I wonder how this makes any sense ...

Sending claude code or codex in a fishing expedition in a dodgy repo or on the web is soooo fun.

Over 4 days codex gpt-5.4-pro created a solution that span close to 13K lines of code. Claude code Opus 4.6 refactored it to 4k5. The runtime & correctness is the same. What's going on. it used to be the other way round.

GPT-5.4-pro is so slow i can never get close to the maximum quota with codex. At 1m TPM I'd need 62k calls to hit the threshold ..

If you want to open a PR for @openclaw don't do it. There's anyway a PR for that in those 5000 existing PRs.

When someone self declares themselves as being kind or empathic toward another, during a circumstance, that's exactly what they're not! They cannot possibly be kind to someone and at the same time their brain goes "Look how much kindness/empathy you have towards X. U cool!" Run

According to OpenRouter Opus 4.5 runs faster on TPU than Nvidia GPUs

No great wisdom has come without a touch of madness.

It's fascinating that @openclaw/@clawdbots used typescript and github.com/badlogic/pi-mono instead of the langchain stack. what a missed opportunity for langchain.