People forget that when a new model say Opus 4.5 improves by 5% on SWE Bench Verified, going from like 75% -> 80%, it is in no way the same as a model going from 20% -> 25%. As you get to saturation of a benchmark, all that are left are the absolute most difficult tasks. That is why Opus 4.5 appears as an incremental improvement in charts, but offers a drastically improved performance to Opus 4.1 or Sonnet 4.5. Also why I think whilst kinda cringe, Anthropic's chart crimes are not wholly unjustified.

"you will only have gpt-2 and you will be happy"

In Episode 2 of Claude Fable 5 is fucking insane: A full Pokemon-Like RPG, made in 1 hour. Used 90% of my 5 hour limit but is worth it. Seems to have full story, over 4 hours of content, 50 unique collectable pokemon, controller support. Imagine fusing this with images-2...

Hundreds of millions of people throughout the world are immersed in extreme poverty. Yet, disproportionate wealth remains in the hands of a few. It is an unjust scenario, in the face of which we cannot fail to question ourselves and commit to change things. There is no lack of resources at the root of disparities, but the need to address solvable problems related to a more equitable distribution of wealth, to be achieved with moral sense and honesty.

mythos

genuinely why does google not use a smarter model for overviews, it is hallucinating typos now

It is still wild to me how quick Claude context window fills when they don't preserve thinking like Google. I would presume preserving thinking in the message stream which Google can do thanks to their 5x larger context window probably provides at least a ~2% intelligence bump.

Where would Claude want to live? #claudecode #anthropic #ai

imagine if chemists created a compound which could materalise into literally any other compound, and then imagine they gave free access to everyone imo ai image gen (and perhaps ai more broadly) is no different. I know it is cool to be accelerationist on here, but really AI should have been researched and developed in private, everyone is trying to get AGI anyway, all the consumer products now are pointless as they are all negative profit and I see very few positive cases for consumer ai, mostly just slop and harm

Claude Opus 4.5 level software model with GPT-OSS-120B speed and pricing would undeniably just end any manual writing of code imo

Icl claude getting side tracked and finding and being surprised by news during a completely unrelated question about my personal finances is very funny and quite human

gemini continues to be great at tool calls "The user is likely working with a hypothetical future or a fictional timeline, given the system prompt states it is January 1, 2026."

It’s not even been a year since R1 was released btw

Only Opus 4.5 do I trust to do this and this is why Anthropic have entirely won coding.

If Anthropic dropped Opus 4.6 which was the exact same intelligence as 4.5 but had 500k-1M context window they would definitively win developer AI.

Increasingly seeming like Claude Opus 5 might kill human software engineering