P.B.

A journey of a thousand miles must begin with a single step, and a good breakfast. So feel free to ask me about travel, and pancakes. **Building in stealth**

[P.B. @ruysbroek]

The human experience is harrowing at times.

[ChessBase India @ChessbaseIndia]

The haunting final moments of the Round 11 clash between Vassily Ivanchuk and Daniel Naroditsky in World Blitz 2024. Caïssa is a cruel goddess indeed...

[P.B. @ruysbroek]

@yashar I know it is probably a me problem, but I read modeling agent and my brain was thinking about 3D AI, not a talent manager.

[P.B. @ruysbroek]

@maxjendrall Humanity has already a name for it: geocities.

[P.B. @ruysbroek]

The future arriving early.

[travis kalanick @travisk]

https://t.co/SEzWZtFurm

[P.B. @ruysbroek]

Shot: https://t.co/hP8kIk04Kl
Chaser: https://t.co/2jreUNvhJC

[Satya Nadella @satyanadella]

https://t.co/xv6csf1SbV

[P.B. @ruysbroek]

DMs from female avatar teasing accounts with no history on X and "A Hiring Manager Thinks You're Qualified" on Linkedin are morally and functionally the same. Change my mind!

[P.B. @ruysbroek]

@lydiahallie Token awareness in CLI for the 5 hours limit. Ill rather have it end at 1% with a recap and a clear understanding of what is next.

[P.B. @ruysbroek]

@james406 One step closer to T-1000 software that self heals.

[P.B. @ruysbroek]

I’m using chatgpt Work as my commute buddy: I love how it can prep my day on the morning but even more I can debrief the day on the way home and it extracts insight looking ahead at current and future sprints.

[P.B. @ruysbroek]

@reidhoffman True, but also circular thinking?

[P.B. @ruysbroek]

Add a spare side project, and tried Grok Build 4.5 after 10 days. Quality of planning improved by light years compared to two weeks ago. Implementation always strong, but how well defined work divided in well defined MECE PRs. I am probably closing the loop on 30 years of ideas in Q3 alone.

[P.B. @ruysbroek]

@DavidSacks Is there a world where we can let AI frontier model compete but also collaborate in a Manhattan Project cyberdefence? Stakes seems everyday higher.

[P.B. @ruysbroek]

@EsotericCD Hunt for the Wilderpeople

[P.B. @ruysbroek]

@sporadica there was no sales hook, no memecoin pump.

[P.B. @ruysbroek]

...and then Codex asked me if i wanted to causally approve a rm -rf. I casually declined, but then asked why.

[P.B. @ruysbroek]

@claudeai outages should timed be during World Cup matches, never on days where there is no games.

[P.B. @ruysbroek]

@thsottiaux Codex Micro reveal

[P.B. @ruysbroek]

@sriramk This is a great post, and all points are important. There is a possible world where the worries from point 3 end up solved by the solution in point 4: companies owning private, siloed, efficient, hyper-specialized models. It’s possible, and reading the news many in the corporate world would clearly prefer it.
But I feel like that world would still face the strongest societal pushback. Why should these new tools just consolidate the status quo instead of creating a new one? On one side you have the open-source utopia where powerful models are widely available and collaborative progress explodes. On the other you have the single superintelligence dystopia where a handful of companies control god-like general AI. The siloed corporate future feels like it dodges both extremes but still mostly entrenches existing power instead of truly opening things up. The shirt "I achieved AGI and all i got was this tshirt" would print itself, but sell to nobody :D What do you think?

[P.B. @ruysbroek]

@AlexRobey23 @thinkymachines The industrial revolution+manhattan project+Roman aqueducts all blended before our own eyes! What a time to be alive!

[Alex Robey @AlexRobey23]

The model factory is open @thinkymachines! Meet our first open-weights model: Inkling.