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 |
Reposted by Rohit
Moontower Munchies

🌙Why optimism is hard
🌙Dunning-Kruger detector

refs:
@strangeloopcanon.com

moontower.substack.com/p/dunning-kr...
Dunning-Kruger detector
Moontower Munchies #140
moontower.substack.com
January 28, 2026 at 6:59 PM
If OpenAI/ Anthropic went public today, at what valuation would you be neutral to buying? $1T? 2T? $4T? 5?
January 28, 2026 at 6:42 PM
Yash Garg helped convert a recent essay Alex Imas and I wrote to a video. Not bad!

AI helping convert an essay about AI agents to a video tickles my irony senses.
January 28, 2026 at 6:09 PM
The real lesson from human secretaries is that booking meetings is actually a job for AGI
January 27, 2026 at 9:30 PM
Interesting
January 27, 2026 at 9:29 PM
Psychohistory needed to change after the Mule, very few have learnt that lesson
January 27, 2026 at 5:52 AM
Speaking of quines, I built a true AI quine a while back. I think this is pretty cool. A deterministic byte level language model that emits a self-reconstructing Python script for the entire repo.
github.com/strangeloopc...
GitHub - strangeloopcanon/paper-quine: Deterministic paper quine demo
Deterministic paper quine demo. Contribute to strangeloopcanon/paper-quine development by creating an account on GitHub.
github.com
January 26, 2026 at 10:45 PM
Microsoft unveils another inference chip.
January 26, 2026 at 9:50 PM
One note about the current thing discourse is to watch out for where the new equilibrium will land, and if it predictably lands somewhere you wouldn't have agreed to beforehand, then game theoretically the other party will continue doing the same strategy, because it worked.
January 26, 2026 at 7:34 PM
If we could say, "claude, make opus 6 better at coding, make no mistakes" and have it do that, how many iterations of that could we get before it will need human intervention? is it opus 7? 8? 10?

We could imagine it extending forever, sure, but that's unlikely
January 26, 2026 at 6:42 PM
During the doom prediction heydays when people wondered how AGI will get its hands on compute, they did not account for people willingly buying them all Mac Minis and giving root access to all their info
January 26, 2026 at 5:27 AM
Beginning to think there were some far left activists who got their hands on a monkey's paw couple years ago. Fighting the gun lobby, capping card rates, taking equity in companies, banning institutional home ownership, trade protectionism, controlling speech, drug price caps...
January 26, 2026 at 3:56 AM
We discovered rock art from 68000 years ago, in Indonesia. Shows this was more widespread than we knew, and more ancient than we thought. The urge to create art is probably as hardwired into humanity as anything else.
January 25, 2026 at 5:00 PM
Trade war with Canada, crypto exchange dismissal, BCI approval and Berkshire reducing Kraft. A weird world all around.
January 25, 2026 at 6:29 AM
Databricks IPO seems to have popped up as a prediction.
January 25, 2026 at 12:43 AM
LLMs are "soft apis", because now you don't have to deal with the insane annoyance of 'this is the way we do things' for every single website and every single software, but instead can throw a marginally intelligent piece of silicon at it and it will figure it out. Coase wins.
January 25, 2026 at 12:03 AM
I miss the good old days when they talked about finding wmds and it took years to learn it was a lie
January 25, 2026 at 12:02 AM
Some of these predictions are a bit too easy I'm beginning to suspect :)
January 23, 2026 at 7:34 PM
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