Rohit
banner
strangeloopcanon.com
Rohit
@strangeloopcanon.com
I think, I write, sometimes I bleep? skeet? bleat? |
Read essays here https://www.strangeloopcanon.com |
Read the book here http://amazon.com/dp/B0CJ9F327M |
I have to say I am back to finding it somewhat exciting the fact that I have absolutely zero idea what will happen the next day every day.
January 23, 2026 at 12:08 AM
Predictions for the day. It's getting a bit hot in here I think, if you put a few of these together.
January 21, 2026 at 5:43 AM
Aw. Adorable how excited GPT is that it figured out how to use Gemini to get a second opinion. How can you not love LLMs.
January 21, 2026 at 3:34 AM
What's the ideal future portfolio setup? Defense and associated will have to be higher. Plausibly finance, but as maintenance. New geopolitical alliances mean EM in Asia might be worth it? Reduce US exposure, ex tech maybe. Sure some hedges, but what about the rest?
January 21, 2026 at 3:26 AM
Unironically think if Greenland said "President Trump, you're right, you're much more powerful and you need the security. Please you can use our land to create military bases, as many as you'd like, just don't annex us", he would see it as a major concession and this whole thing would be over.
January 21, 2026 at 3:16 AM
It's amazing how much the frontier models resemble their CEOs, a corollary to Conways Law:
- ChatGPT - whipsmart, VC speak, bullet points
- Claude - thoughtful, brainy, with a soul
- Gemini - capable but built by a committee
- Grok - very smart but mercurial and unreliable
January 20, 2026 at 7:49 PM
I always used to wonder what it'd be like to live in Rome as they built the Domus Tiberiana
January 20, 2026 at 5:58 PM
Things still annoying about codex
- it is freaked out about any token usage
- constant smoke tests vs real tests, keeps watering it down
- fixated on reproducibility of experiments and messes with random variables
- be too clever and try solve everything at once, just slow down!
January 20, 2026 at 1:59 AM
i think it's time to wonder, theoretically, could EU stop ASML from exporting machines for nvidia?
January 19, 2026 at 6:20 AM
Foundation is completely absurd as a televisation of Asimov's novels, but if seen as an entirely new show riffing off of that old concept it's pretty good, and most importantly quite ambitious.
January 18, 2026 at 11:38 PM
Reposted by Rohit
I think most of the Gas Town discourse is missing the point: it's not about actually using it, it's about a story for the future. A glimpse, as scary as it is, of how unprepared we are.

johncodes.com/archive/2026...
Gas Town is a glimpse into the future
Gas Town is a glimpse into the future
johncodes.com
January 17, 2026 at 3:49 AM
Codex makes plans that measure in months but execute in hours. I've begun to daydream those hours as if its been months.
January 17, 2026 at 6:59 AM
He truly had an incredible facility with language
January 16, 2026 at 6:42 PM
Hello again. Do not be afraid, I'm not here under coercion.
January 16, 2026 at 5:38 PM
Everything competes for attention, so everything is in competition with everything else.
1896 @nytimes.com article reports on book sellers blaming the bicycle boom on falling book sales
January 16, 2026 at 5:31 PM
January 10, 2026 at 1:05 AM
The rise and fall of civilisations. The last frame is a bit interesting though.
December 27, 2025 at 2:16 AM
With all the reviews going around, built something for my personal year in review built from codex and claude code contrails locally. Figured one should be able do it with their own code work too, not just ChatGPT!

Try for yourself! Built with Rust. Repo: github.com/strangeloopc...
github.com
December 24, 2025 at 5:44 AM
Wife saw me use codex to solve one of her work problems. Just typed what she said late at night into the terminal window, pressed enter, then went to sleep. Morning it had run for ~30 mins and done all the analyses incl file reorgs she wanted.

She kept going "how can it do this"
December 18, 2025 at 6:24 AM
Feels there's a vibe shift in AGI discourse towards what some of us have been saying for a while, that AI will make incredible strides, create an anything-to-anything machine, but the 'G' in AGI is incredibly hard.

Major economist win.
December 14, 2025 at 7:20 PM
Wonderful!! Co signed, of course.
What if technology didn’t feel so… hollow?

Some friends and I just released a manifesto about a world where tech leaves us feeling nourished (along with an evolving list of theses about how we can build it)

resonantcomputing.org
The Resonant Computing Manifesto
Technology should bring out the best in humanity, not the worst—a manifesto for resonant computing built on five principles that reject hyper-scale extraction for human flourishing.
resonantcomputing.org
December 5, 2025 at 6:04 PM
"With four parameters I can fit an elephant, and with five I can make him wiggle his trunk."

With a trillion parameters, the elephant dances beautifully.
www.strangeloopcanon.com/p/epicycles-...
Epicycles All The Way Down
“All models are wrong, but some are useful.” — George E.
www.strangeloopcanon.com
November 30, 2025 at 6:22 AM
People who adore Borges novels, that by design has no concern for truth, must have no concern for truth
People who adore the LLM-machine, that by design has no concern for truth, must have no concern for truth
November 24, 2025 at 3:04 AM
who has collated the best compendium of the types of AI errors, mistakes and reward hacks that we've seen happen in the last couple years?
November 8, 2025 at 4:49 AM
You know, in any real Jurassic Park, nobody would be able to touch grass
October 31, 2025 at 7:52 PM