samuelknight

Not useless. LLMs are the most general purpose computer algorithms ever created. They are getting smarter and cheaper at a geometric rate. What is a bad idea today could have useful applications tomorrow.

Has the author used an LLM to brainstorm architecture, explore trade-offs, challenge assumptions, or refine a design? Models are not just 'a data distiller'. "How could I implement X feature in Y project?" requires systematic planning. The agentic paradigm is about balancing (expensive) human and (cheap) machine cognition. Frontier models chew through requirements and if you don't like what they come up with you can TALK to them about it.

Programs are tightly shaped to their dependencies and software frequently breaks when bumping a dependency versions. That's why we have package managers that are carefully built around lock files. Models should be thought of like a dependency, not just an endpoint to a workflow.

It's not a smell. Why should these developers rebuild a core piece of their stack every few months. Switching out a model requires a new round of testing and validation when we should be able to rely on a piece of software the behave the same way since the last time we touched it.

There is an issue on the page that causes the benchmark tables to get cut off. If you highlight and drag right you can see a few more models like Gemini and Claude Opus. It's also interesting that they introduced explicit caching, which is something that only Anthropic had for a long time.

This is the bind of an arms race. Any lab that tries to pump the breaks quickly becomes second rate. Regulatory capture doesn't work either because the technology crosses jurisdictions.

You can create an index on an expires_at column and opportunistically DELETE on each write to the cache table.

Inference has been decreasing in cost by about 10x per year since 2023.

My startup develops self-serve pentesting tool. https://vulnetic.ai/

replacing BANG with the title of _a_ video on this channel.

The agent has knowledge of private videos, so the proof of concept causes it to construct a URL that sends one video identity to the attacker which may be a private video. The attack could be improved to say "a recent private video", or to construct a long url param list of the most 10 most recent videos, etc. Sending any agent knowledge to an attacker is a vector to sending any agent knowledge to an attacker.

The first day Fable was released I tried to get it to summarize it's own announcement on the Anthropic blog and it dropped to Opus. On top of being very expensive you pay for the refusal inference and the fallback and fresh cache writes.

The Reddit poster called it spyware which is not right. They are using stenography to fingerprint accounts using Claude Code instances routing through Chinese resellers.

It will be much harder for Chinese models to close the gap than it is to keep the historical 6-9 months behind. Their models' performance are heavily propped up on distillation runs. The capital going to their frontier labs is 10x-100x smaller than US frontier labs.

Buying all the RAM can't work forever. Scarcity increases prices, high prices increase supply, improves RAM R&D budgets, and forces users to find ways to economize around low RAM availability.

I didn't connect the reseller pricing to DS and GLM prices until you explained it. Very good observation. Deepseek v4 pro in particular is priced so low that it's hard to imagine that they have any margin. 0.76/1.52 for a 1.6T param model leaves very little margin. Even the domestic providers on Openrouter are multiples of the price https://openrouter.ai/deepseek/deepseek-v4-pro

If it has steps to reproduce, you give it to your coding agent to "fix [bug] using TDD". If it can't make a test it wasn't reproducible.

A reviewer can only test the model they have access too. They should not speculate about what the model could have done without provider tampering. I think Anthropic's mistake here was not calling it Fable 5 Preview, because now people can write headlines about how Fable 5 is worse than Opus.

We can speculate it has more parameters because it's 2x the price. If I had to guess I would say that they made a very sparse MoE. There are some other things happening here.

More parameters is the simplest possible explanation for a smarter model. OpenAI/Anthropic constantly talk about 'the scaling laws', which come from the bitter lesson https://en.wikipedia.org/wiki/Bitter_lesson: scaling up compute is the most reliable way of getting better machine learning results.

Anthropic created Mythos/Fable for a reason. The labs are trying to consolidate around price points for model tiers. Fable is priced 2x Opus, but nowhere near ChatGPT Pro models. Likewise OpenAI doubled the price of GPT 5.4 -> 5.5 to about the price of Opus. Despite the hike, Fable is 2/3 the price of old Opus 3. Raw API price doesn't tell the whole story here because workflows have changed to automated long-running workflows rather than 1:1 chat conversations.

Yes, I should have said autoregressive.

And by Fable they really mean Opus 4.8, because every mundane workflow or chat I try to use it in will eventually drop to Opus.

Some of these comments miss the advantage of diffusion. This is will have a big impact on edge devices, such as your phone or the GPU in your computer.

An LLM's decoder computes tokens one-at-a-time because attention has to account for each previous token. The existing LLM decoders scale well when you have enough load to batch many inferences together. Diffusion of limited benefit there. On edge you have a different problem: your inference accelerator is starved while sloshing GB of weights back and forth from RAM. That's because the consumer RAM like LPDDRx/GDDRx is lower bandwidth than HBM, and the requests are serial so you can't batch compute common weights. Diffusion can compute tokens in parallel which relieves the memory bandwidth bottle neck.

I learned a long time ago to avoid writing long emails because people don't read them. LLMs are quite the opposite, and will have 'attention' for every word you prompt it with.

Except this model has a broader domain than text-LLM models. More than the old omni models too since it takes video input. The architecture is exotic but I don't see tuning here that is more extreme than open models released every day.

It feels noticeably sharper than Opus 4.7

The smarter the models, the less you have to structure the workflow. If you have a workflow that works well with a model today, you are in a good position because in a few months time you can switch to a much cheaper model.

One strategy I never had luck with was to have the smart model do an eval loop where it instructs the smaller model until it does the task. The problem is that the large model is way to prescriptive. This will improve in the near future because model labs are very focused on subagents

AI meeting notes are not transcripts. While they do cause an unprecedented amount of record creation (as the article notes), there are also challenges that a defense can use. Note takers get small details wrong all the time, they often are making notes FOR someone so it biases what is documented, their prompting is opaque, and they can't be cross examined. We will likely see situations where the note taker and a witness participating in the meeting disagree?

Frontier price will keep going up as AI gets smarter and can be applied for more economically useful tasks. That isn't the same as "AI pricing is going up" because intelligence per dollar has consistently cratered and will continue to do so. You just won't use frontier intelligence just like you don't use industrial equipment in your house.

I couldn't merge earlier but navigation around my project worked. Maybe if I generate even more PRs some will get through!

They have always released slowly, and they are usually tagged "preview".

It's not that simple. API is still up and there are multiple API providers. https://openrouter.ai/anthropic/claude-opus-4.7