Hacker News Reader

simonwon GPT-6.1 Sol replaces GPT-6 Sol after just 7 days, with near-Astra intelligence

It makes sense for them to sit on it until they've finished testing it. More powerful but also more likely to delete all your email by mistake = you shouldn't release it yet.

simonwon Livenerf: Has Opus 5.5 been nerfed yet?

> Anthropic has admitted to nerfing in the past

Where?

simonwon GPT 6.1 Sol: Near-Astra intelligence for a fifth of the price

Sorry about that, markdown bug, now fixed.

simonwon DevDay 2026 Recap

Yeah, the front dozen rows were reserved for OpenAI employees.

simonwon GPT 6.1 Sol: Near-Astra intelligence for a fifth of the price

I'm a bit late with the pelicans because I was live-blogging the keynote: https://simonwillison.net/2026/Sep/29/openai-devday-2026-liv...

Here they are for GPT-6.1-Sol: https://tools.simonwillison.net/markdown-svg-renderer?url=ht...

They're not notably different from the GPT-6 family pelicans: https://static.simonwillison.net/static/2026/gpt-pelicans-gr...

simonwon Anthropic's IPO prospectus shows AI vision, surging costs

Yeah, I agree with all of that.

simonwon Anthropic's IPO prospectus shows AI vision, surging costs

If Uber think there's no ROI, why do they allow each of their employees to spend $1,500 per month per tool?

Shouldn't they have set that per-employee budget to zero instead?

(I dug up the original source for that Uber doubts the ROI story a few months ago, it's a lot weaker than the headlines about it suggested: https://simonwillison.net/2026/May/27/product-market-fit/#th...)

simonwon Anthropic's IPO prospectus shows AI vision, surging costs

> If they had a trillion dollars worth of revenue, it wouldn't mean jack shit if they had a trillion + 1 in losses.

I don't understand that argument.

If a company has a trillion dollars in revenue, even if they are losing money hand-over-fist, that still means they have convinced other companies to cough up a trillion dollars for what they are selling. That's a big deal!

The only case that isn't impressive is if they are literally selling dollar bills for 90 cents.

You can argue that Anthropic are subsidizing their tokens all you want, but since as a customer you can't just turn around and sell a token yourself for more than you paid for it that's still not a good argument for dismissing the amount people are willing to spend.

simonwon Anthropic's IPO prospectus shows AI vision, surging costs

https://www.anthropic.com/claude-opus-5-5

"Opus 5.5 requires less compute to serve than Opus 5, and its pricing reflects that. Our tests show that at default settings it will cost 40% less than Opus 5 on typical workloads."

https://www.anthropic.com/claude-sonnet-5-5

"Sonnet 5.5 requires fewer tokens per task than Sonnet 5, so it’s less expensive to run. It also generates output 30%+ faster"

OpenAI have been achieving even more impressive optimizations, hence why GPT-6 Sol and GPT-6 Luna are half the price of their 5.6 equivalents.

simonwon Anthropic's IPO prospectus shows AI vision, surging costs

I assembled some of those numbers in May, when they claimed they had grown from $9bn in annualized revenue in December 2025 to $47bn in early May. https://simonwillison.net/2026/May/29/anthropic/

Since then they've reported $65bn in annualized revenue by July: https://simonwillison.net/2026/Aug/23/anthropics-best-ai-mod...

And sure, they might be lying about those figures - but if they are, that's investor fraud, and they'll be in hot water with the SEC when they try to IPO. I don't think they are lying about the figures.

If I had numbers on their cost of revenue I would share those. As it stands I'm going to have to wait for either more leaks or their S-1.

simonwon Anthropic's IPO prospectus shows AI vision, surging costs

We've also heard plenty of stories from companies like Uber about a need to set a limit on their token spending because they were spending so much on Anthropic. That's a solid sign that the revenue numbers are real.

simonwon Anthropic's IPO prospectus shows AI vision, surging costs

But that was in 2025. It's not surprising to hear that in 2025 they had two customers responsible for a quarter of their revenue (those companies were Cursor and GitHub Copilot btw) - they hadn't yet sold Claude Code packages to vast numbers of large companies.

simonwon Sonnet 5.5

Linking to https://tools.simonwillison.net/markdown-svg-renderer?url=ht... should be pretty inoffensive (I started habitually linking to that after people kept complaining about linking to my blog) - that page renders Markdown with SVG embedded in it, but doesn't link to the rest of my site at all.

simonwon Anthropic's IPO prospectus shows AI vision, surging costs

I'm confused.

The numbers that Reuters describe in this article were entirely for 2025.

It's been well documented that Anthropic's revenue growth in 2026 has been enormous. This was the year of coding agents, and tokenmaxxing, and companies blowing enormous amounts of money on AI thanks to coding agent users spending hundreds (or thousands) of dollars a day.

Given that, I don't understand why the article and the headline are based exclusively on those 2025 numbers, with not so much as a hint to the reader that there are figures from the past 9 months that aren't covered by the documents Reuters saw.

Losing $8bn in 2025 isn't particularly notable if you've made ~10x that amount of revenue in 2026.

(How much did they lose in 2026? Wouldn't we love to know that!)

Is this just a thing with leaked IPO prospectuses and coverage of them?

simonwon Sonnet 5.5

Isn't posting any comment on a forum like Hacker News "attention seeking behavior"?

simonwon Sonnet 5.5

You can click the little [-] icon next to the post to collapse the entire sub-thread. I do that all the time.

simonwon Sonnet 5.5

Mainly because they're funny, but it's also because I try pretty hard to make the comment more interesting than just "here's a pelican". In this case I used the pelicans to talk about the 128,000 token limit bug at "max" and share comparative pricing.

In the GPT-6 comment I included full visual comparison grids: https://news.ycombinator.com/item?id=49805509#49806126

For DeepSeek v4.1 Flash I identified that the OpenRouter reasoning levels are mapped to a smaller set of levels for that model: https://news.ycombinator.com/item?id=49639090#49645591

simonwon Sonnet 5.5

Yes, the OpenAI GPT-6 Astra limit is 128,000 as well: https://developers.openai.com/api/docs/models/gpt-6-astra

Gemini 3.8 Flash is 65,536 https://ai.google.dev/gemini-api/docs/models/gemini-3.8-flas...

simonwon Sonnet 5.5

I think this is a bug. I've not seen this problem from any of the other frontier models.

simonwon Sonnet 5.5

It's the output token limit, which has been 128,000 for Claude models for quite a while note

simonwon Sonnet 5.5

Pelicans. Sonnet 5.5 has the same problem as Opus 5.5: on "max" thinking effort it burned through 128,000 thinking tokens (taking 15 minutes to do that) and ran out before it had produced the final SVG.

https://tools.simonwillison.net/markdown-svg-renderer?url=ht...

Here's how the thinking effort levels compare:

  low
  27 input, 1,623 output, thinking_tokens: 0
  1.6284
  Duration: 10138ms (10s)
  
  medium
  27 input, 1,796 output, thinking_tokens: 0
  1.7914 cents
  Duration: 11266ms (11s)

  high
  27 input, 2,334 output, thinking_tokens: 745
  2.3394 cents
  Duration: 17376ms (17s)

  xhigh
  27 input, 5,730 output, thinking_tokens: 2535
  5.7354 cents
  Duration: 41882ms (41s)

  max (failed to return response)
  27 input, 128,000 output, thinking_tokens: 128000
  $1.28
  Duration: 940617ms (15m 40s)

Low and medium both used 0 thinking tokens.

simonwon The problem is not AI code, but not knowing about system architecture or intent

Delving into the code is not the same thing as reading every line.

simonwon The problem is not the AI code, but nobody knows anything anymore

My hope/hunch is that the kids will be alright. Not learning effectively is a choice: if you want to get good, the paths to getting good are all still available to you.

We have never had as abundant a supply of tools to help us learn our craft. I expect that many people will thrive.

People who are a bit lazy and prone to cheating will be able to hurt themselves even more.

simonwon The problem is not AI code, but not knowing about system architecture or intent

If you don't understand how your system works, your ability to make good decisions about future work on that system quickly degrades.

I don't think you need to review every line of code, but you absolutely do need to be able to describe how the system works and its high level structure.

As is so often the case with coding agents, having experience as a tech lead or engineering manager really helps here. You are responsible for a large system that has been worked on by multiple different collaborator (both human and agentic). You need to be able to make smart, informed decisions about that system, and talk with credibility to other stakeholders about what it can and cannot do and sensible next steps for the project.

simonwon Jensen Huang says AI distillation is 'competition.'

Jensen Huang sells GPUs to the highest bidder.

simonwon Prompting Claude Opus 5.5

That "mark pasted text" thing is interesting: https://platform.claude.com/docs/en/build-with-claude/prompt...

  Summarize the main complaints in this thread.
  
  <pasted_content id="ab12">
  ...text the user pasted...
  </pasted_content id="ab12">

Where those IDs are randomly generated and unknown to the user, and the model is told to use that markup to help avoid it suffering prompt injection attacks.

In the past I've been very skeptical of this kind of protection. Anthropic have clearly trained their models for this though, so maybe Opus 5.5 is smart enough for this to work?

Will be interesting to see if minds more devious than mine can break it.

simonwon What I did at Recurse Center

> Has there been a business started based on work done there?

Greg Brockman went through Recurse in Summer of 2015, and co-founded OpenAI in December of 2015.

simonwon S3 Is the Future, S3 Is the Past

One thing I find notable about S3 today is that, while it used to drop in price reasonably often, there hasn't been a price drop in a full decade:

  2006-03-14  $0.150/GB-month
  2010-11-01  $0.140/GB-month
  2012-02-01  $0.125/GB-month
  2012-12-01  $0.095/GB-month
  2014-02-01  $0.085/GB-month
  2014-04-01  $0.030/GB-month
  2016-12-01  $0.023/GB-month

Today it's still $0.023/GB-month.

simonwon Jevmem – automatic project memory for Claude Code, built on Jev

Yeah, I've started dialing back my use of AI for READMEs because of this.

My previous rule was that I never use AI for writing that expresses my own opinions or tries to be convincing (anything on my blog for example) but I'll let it do technical documentation.

The top of a README is about convincing and explaining why I built something though, which means it should fit my no-AI policy after all.

simonwon Opus 5.5 is good at explainer videos

I trust my workflow a tiny bit more, because I don't know how that one works.

Plus having the video file means I can extract frames as images at specific timestamps.

Homely though the ask feature is probably fit for purpose, at least on YouTube.