- Serial (produces an incredibly exciting response which ends in a cliffhanger that withholds the answer)
- Prequel (instead of responding, it provides the full backstory leading up to your question)
- Yarn (maximizes output tokens by taking a long winding route to your answer)
- Head Canon (answers using its own entertainingly weird theories about the input)
- Overstory (your answer is interwoven with the answers from eight other users into a larger and deeply intertwined meta-answer)
- Oeuvre (for every question, it produces a diverse but cohesive body of work across a variety of mediums, each one a heartbreaking masterpiece in its own right)
Thankyou for sharing this. I've been reading the culture series over the last few months and this beautifully articulates ideas that I've not been able to.
I sometimes wonder what goes through designers heads.
The article is surely good. But it's long form and I need to pay attention to it in order to appreciate it.
Yet they decided to add an obnoxious banner to the view which cannot be hidden and keeps distracting me every few sentences by being significantly more attention grabbing then the article.
So I stopped reading. I would've likely read to the end and consequently been more willing to spend money, but I never got to that point.
Yes I have ADHD, and I'm sure other people that don't have it don't have such an issue. But to me the design for that element is in direct contrast to the content being served. If the article was short form it'd be able to ignore it, but with it being long form, I cannot and thus my experience was actually negative.
The Dario Legendarium is definitely going to be a fun piece of work for historians to interrogate as to the origin of this age. The relative personalities of the orgs involved show somewhat in their naming: {o1,4o,3.5,5.5} vs. {Haiku,Sonnet,Opus} vs. {3.1, 3.5-pro, 3.5-flash, 3.5-flash-lite} vs. {3.7, 3.7-plus, 3.7-max}
It's a pity that Samsung isn't in the mix. I would have liked to have used a model named Samsung Galaxy S10+ 5G Lite.
Surely it would be Tàiyī (Supreme One) or Tiānjī (Celestial Mechanism) in keeping with Tiāngōng (Celestial Palace) for their space station. Though naming a small model Tàiyī would sound stupid haha! Xiǎoyī!
Anthropic intentionally degrading responses shows you they're also focusing on the wrong things. Rather than producing the best model possible, they're nerfing their models' capabilities.
"Fable" is not the name I would have chosen for a product that has to argue for the fact of its own economic viability but it is the correct mood. Perhaps Anthropic is trolling Zitron.
I miss the days when the dropdown menu (in their consumer product with a billion users) asked me if I wanted to use o3, 4o, 4o-mini, o4-mini, gpt-4.1, gpt-4.1-mini, or gpt-4.5 (Research Preview).
Tractatus - highly aware of its limitations and refuses to speak about too many topics. Really likes numbering statements. Unusable due to a context window bug which is patched after considerable delay.
This is a well executed joke, from the design to the shortness to the being on point to the abrupt punchlines when one looks at the lower ones and it also vibes well with what we all were thinking anyway, and the unexpected seriousness of the title.
I like "Proverb" as smaller than Haiku too, Aphorism is also good. But seriously I want Anthropic to up its small model game. Haiku is not competitive, Deepseek v4 flash outperforms my uses for about $0.10 / $0.20. Whereas Haiku 4.5 is $1/$5.
IMO Anthropic should just play the game at all the price tiers because it otherwise forces people to go elsewhere. I would probably pay for a "Proverb"/"Aphorism" class model that was worse than Deepseek at the same price just to stay in the ecosystem, if given the option.
(Note: I also see Google seem to make the same mistake, they actually do have competitive models in Gemma family but they don't make them available via the API. So there may be some reason for this.)
- Serial (produces an incredibly exciting response which ends in a cliffhanger that withholds the answer)
- Prequel (instead of responding, it provides the full backstory leading up to your question)
- Yarn (maximizes output tokens by taking a long winding route to your answer)
- Head Canon (answers using its own entertainingly weird theories about the input)
- Overstory (your answer is interwoven with the answers from eight other users into a larger and deeply intertwined meta-answer)
- Oeuvre (for every question, it produces a diverse but cohesive body of work across a variety of mediums, each one a heartbreaking masterpiece in its own right)