I think people are putting far too much weight on the Sonnet 5 benchmarks when the vibes (aside from token efficiency) seem pretty good.
We're all used to Open Weight models bench-maxing to show competitiveness with frontier closed models, but this might be causing some people to misjudge the baseline intelligence of models that don't train so hard against the common benchmarks.
Sonnet 5's scores on newer, higher signal, benchmarks actually seem really competitive with the (currently available) frontier, although I'm still waiting on DeepSWE.
nitter.gq/cognition/status/20720…
My impression of Sonnet 5 is that it's primarily a response to Open Weight models and the Fable 5/GPT-5.6 situation. Its intelligence/pricing make it competitive against GLM 5.2 without exceeding Opus 4.8's capabilities, avoiding questions about restricting frontier intelligence
Some are saying he's the best to ever do it
Mamdani brought along the fake Daily News front page as a gift for Trump to cajole him into providing funding for a long-stalled plan to build 12,000 apartments on top of Sunnyside Yards, per City Hall & sources.
More on Mamdani's latest meet with Trump: politico.com/news/2026/02/26…
We're now talking about which presidential actions have absolute immunity versus which *might* be criminally prosecutable. Unbelievable that the United States has come to this.
This quoted post is unavailable.
Chase C retweeted
But they're BOTH so OLD! Gandalf can barely remember his passwords and even forgot that people called him Gandalf (yikes), while Saruman betrayed the Free Peoples of middle-earth and led an insurrection that stormed Helm's Deep. You can see how they're equally bad choices.
ALT Gandalf the Grey
ALT Saruman
Chase C retweeted
I still haven't heard a good answer to this question, on or off the podcast.
AI researchers often tell me, "Don't worry bout it, scale solves this."
But what is the rebuttal to someone who argues that this indicates a fundamental limitation?
Chase C retweeted
Wrong on every single allegation. (1) What is going to Ukraine will mainly be stored equipment paid for years ago that would never have been used by a U.S. soldier again; (2) the "cost" is spread out over multiple years, as the accounting entry based on book value (higher than marketr value) is used to produce upgraded weapons in US (that will never go to Ukraine); (3) the annual amount is a trivial part of federal spending - less than 0.07% of the federal budget. Sacks and others deliberately use ther multiple year number as if it was annual and they ignore the relevant comparisons;
(4) The stealing quote does not refer to U.S. aid (another deliberate misrepresentation) but refers to corrpution that is already subject to criminal enforcement involving non-aid funds and transactions. U.S. aid is audited and reports go to Congress: no reported issues even from anti-aid reps who sit on committees that get the reports.
(5) Sacks knows nothing about the military and Ben Hodges and other experts have stated a Ukraine victory is "inevitable." They have destroyed over half of the whole Russian army and need a bit more help to drive the rest out of Ukraine (while minimizing their loss of life).
The question remains is why Sacks repeatedly spends time on this issue, lying and obsfucating the facts. Is it another 15 minutes of fame or pecuniary?
Chase C retweeted
ai influencers be like
Waluigis in superposition were just the beginning. This week alone 37 INSANE hyperstitial simulacra were born from our accelerating Molochian memepool!
Here’s what YOU need to know to avoid 10↑↑10 years of punishment under Rococo Basilisk’s rule👇🧵
Chase C retweeted
The solutions exist, we just need the collective will to implement them.
It's April 2026. I wake up in the morning and check Hacker News. "Hundreds of Starlink satellites burn up in the atmosphere." 🧵
