HNHacker News
TopNewBestAskShowJobs

grey-area

24,562 karma · joined April 12, 2012

submissionscomments
grey-area··on Nvidia was in 1 in 11 startup rounds over $100M last year
I’m not sure what about the facts of Nvidia subsidising many of its customers you think is a conspiracy theory.

If we do see a crash and they are sued by shareholders for fraud, perhaps you will reconsider your position.

grey-area··on A year after the 100% AI mandate nobody understands our code
Incumbents are in the worst possible place to start chasing 30k commits a day, it is not going to end well if that’s what they’re doing.

For startups it makes some kind of sense to take risks in this way, for incumbents it makes no sense.

grey-area··on CoreWeave bond yield tops 16% as AI credit stress rises
There are weeks when decades happen.
grey-area··on CoreWeave bond yield tops 16% as AI credit stress rises
A few problems for Anthropic or OpenAI:

1. Model Build/Training costs are escalating if anything

2. They are in an arms race with competitors and can't afford to stop building new models - if they lose the perception of leading, they lose lots of customers overnight.

3. Inference margin is under attack from more specialised models and open weights models.*

4. They have no moat given how simple the interface to the products is - a bunch of text files and text prompts, and little prospect of building a moat.

So those all run counter the possibility of coasting on previous research. To me the most promising avenue for making money for these companies is to replace Google, not AGI or taking over the world's knowledge work. That would mean they are overvalued. Even coasting on previous research, how long would it take to pay off the money they have spent?

* https://strandsagents.com/blog/amazon-saves-millions-replaci...

grey-area··on CoreWeave bond yield tops 16% as AI credit stress rises
Next year is a long way away.
grey-area··on Claude Fable 5.2 and Mythos 5.2
People prefer a chat interface to a search box.

OpenAI is coming for google’s cash cow ads business (which is already vulnerable because of their sharp practice).

grey-area··on Show HN: STE80 Humanizer – de-AI text by writing 80% of the way to ASD-STE100
Well it matters if you’re trying to adjust writing quality higher yet giving it terrible writing in the context.

The more slop you have in your context the sloppier the text you get out will be.

grey-area··on Show HN: STE80 Humanizer – de-AI text by writing 80% of the way to ASD-STE100
Yes it absolutely is slop, full of AI tells. Why do we need all this ridiculous ceremony for a text file of instructions anyway?

If I want to control how it writes I just put a style_guide.md in the context.

grey-area··on A year after the 100% AI mandate nobody understands our code
It doesn't have to be this way. LLMs can be quite helpful in analysis and small coding tasks. I understand everyone doesn't control their environment, but surely there are some sane places out there?

We don't have try to use them as 'agents' to write and understand all the code for us, frankly I wouldn't trust them to do so, they make too many mistakes and foolish decisions about architecture.

grey-area··on Claude Fable 5.2 scores 65 on Artificial Analysis, highest so far
I P O
grey-area··on Claude Fable 5.2 and Mythos 5.2
A growing percentage of the users on the internet are not human.
grey-area··on Nvidia was in 1 in 11 startup rounds over $100M last year
In this case (gpu provider funding many startups to buy their gpus) it seems bad to me. They have deals with so many customers it looks like collusion to keep prices high on both sides.
grey-area··on Burying the Evidence: nine out of every 10 time capsules go missing
We only have 2 books out of around 16 from Homer, so a lot has been lost already.
grey-area··on Nvidia was in 1 in 11 startup rounds over $100M last year
So you’re arguing it is circular but it’s only a little bit of money compared to Nvidia revenue so it doesn’t count?
grey-area··on Nvidia was in 1 in 11 startup rounds over $100M last year
Also evidence that Nvidia is funding its own customers and artificially boosting demand. What could possibly go wrong?
grey-area··on I'm not convinced the AI bubble exists
For certain values of temporarily. On human time scales, we do see bear markets which last a decade or longer, and that can have a serious impact on someone's earning ability or retirement.

Markets do include significant risk and we would do well to remember that.

grey-area··on I'm not convinced the AI bubble exists
It does very much vary with the domain though. I haven't personally noticed much improvement, they're still useful but unreliable, in my view the idea of agents, harnesses etc has been a distraction and functions to cover up the lack of progress in the underlying models.

But everyone has a different view of this, it'll be interesting to see how it plays out.

grey-area··on I'm not convinced the AI bubble exists
And yet we have to contrast that with reports like this of model capability falling recently:

https://news.ycombinator.com/item?id=50422069

It's really very hard to pin down progress in this domain, companies are definitely targeting benchmarks.

One bit of evidence I would look at is the maths results this last year which have been impressive, that was a huge expensive brute force effort, but it is still good progress.

grey-area··on Is Opus 5.5 nerfed? Thinking patterns changed on November 3
These machines are inherently unreliable.
grey-area··on I'm not convinced the AI bubble exists
Unfortunately, they are, the market (even the global market) is heavily concentrated in just a few US listed companies, and other indexes closely track US indexes. A crisis in AI would impact everyone.

For proof, look at the fall in VWRL last week correlated with concerns over AI which hit nasdaq.

grey-area··on I'm not convinced the AI bubble exists
Have the models become significantly more capable in the last 6 months? I'm not convinced we're still seeing massive improvements as we saw at the start. The output certainly isn't getting better, the software these companies produce is actually pretty buggy.
grey-area··on Agent memory needs its OpenTelemetry moment
There are significant problems with text files for memory though - context length, memories conflicting with base weights and with each other. Memories not always being considered and forgotten because they’re in context rather than from training weights.
grey-area··on Claude Code's "Don't Ask Again" exposes passwords in 1,122 public GitHub files
Text in a model context is not a reliable constraint.
grey-area··on Agent memory needs its OpenTelemetry moment
I imagine you probably wrote an outline with the help of Claude on this topic, then got it to generate the article including illustrations based on that, gave it a look over then pressed publish? Did you write any substantial parts of this yourself?
grey-area··on Agent memory needs its OpenTelemetry moment
Should we even be storing 'memories' in text files at all?

Imagine how humans would function if we carried our memories around in overflowing notepads and had to consult them every time we wanted to know a fact or remember a person we had met, it would be exhausting and inefficient.

grey-area··on Agent memory needs its OpenTelemetry moment
At the moment it's a good proxy for quality and intention. IF it was mostly generated, it's probably mostly meandering bullshit which doesn't quite make sense.

Yes the style is annoying and bombastic, but the objection goes deeper than that; it's annoying to have someone generate text with an LLM from a small outline or prompt, not even edit it for sense, and dump it on readers.

grey-area··on Agent memory needs its OpenTelemetry moment
Often people don't even read the article, they just see a headline they agree with and upvote it.

I usually read the comments before deciding if the article is worth my time.

grey-area··on Show HN: Coldframe, a honeypot that poses as a supermarket fridge controller
Of course they did. That text is Claude.
grey-area··on Claude Code's "Don't Ask Again" exposes passwords in 1,122 public GitHub files
That’s good to hear, it’s a much better approach than expecting humans to read commands and approve them or just trust that the LLM can set the proper horizons for its world itself while working.
grey-area··on The Components of Agent Memory
The way we store memories is to adjust the weights in our brain (as far as we know). The way models store memories seems to be to write them down on a scratchpad someplace (details vary, but roughly this is what they're doing).

The article explicitly excludes this from consideration early on though:

By “agent memory,” we mean knowledge that Claude gains from experience...saved as context it can read later. It excludes other ways models can learn or manage context, such as changes to their weights

Those seem fundamentally different processes, and adjusting weights does seem a much better idea after all, but then in that world everyone has a different AI and how do you apply updates? For a company like Anthropic this is of course a big problem and that is probably why they don't like that route. Could we store weight deltas or something similar and reapply them over a different model?

← PreviousPage 2 of 34Next →