pmxi

The key claim here is, paraphrased: “LLM coding assistants” cannot “help someone write code better or faster.”

This is not true.

This is expected behavior; the Codex app has been rebranded as the ChatGPT app.

This catches Claude resellers. Meaning companies who proxy Claude traffic for users in, say, China.

https://www.chinatalk.media/p/how-to-buy-cheap-claude-tokens...

This has been a niche where Mistral has actually been successful. Btw, Hindi and Japanese are bucketed in "Rare Languages," which is odd.

PG acknowledged that growth is limited by market size. Markets to support "avogadrillionaire" level wealth may not current exist.

However, there obviously exist markets to support billionaire level wealth, as evidenced by the at least 30 billionaires produced by YC, and the many others.

I had Claude straighten it out:

  Model           In     Out    BIn    BOut
  Haiku 4.5   $ 1.00  $ 5.00  $0.50  $ 2.50
  Sonnet 4.6  $ 3.00  $15.00  $1.50  $ 7.50
  Opus 4.7    $ 5.00  $25.00  $2.50  $12.50
  Opus 4.8    $ 5.00  $25.00  $2.50  $12.50
  Fable 5     $10.00  $50.00  $5.00  $25.00

It’s near the frontier meaning it’s the best intelligence for the price.

It’s not even close to frontier meaning it’s the best intelligence.

This will load the PR easily: https://diffshub.com/oven-sh/bun/pull/30412

The first part, describes a decision to ask AI over calling a good friend and domain expert. Typically the domain expert is not an option. (looking at my last 10 Claude chats, I knew a domain expert for one of them).

The second part assumes AI writing can substitute mastery of writing, and simultaneously, not be good enough.

We have, for a several thousand years, continued to develop labor and cost saving technology. We have been OK thus far.

Yeah. They are aware: "Users will find Opus 4.8 to be a modest but tangible improvement on its predecessor."

In the "What's next?" section, "There’s still more to be done: we’re working on developing and releasing models that provide many of the same capabilities as Opus at a lower cost."

I wanted to ask someone about the firepass subscription. It implies unlimited usage - sounds too good to be true. Is it? Can you leave your agent running all day?

These CLIs are for AI agents. If I have a CLI to WhatsApp, then I can direct an agent (such as OpenClaw) to manage my messages for me.

I didn't like them either. Thankfully, they can easily be disabled. See my config: https://github.com/pmxi/dotfiles/blob/e779c5921fbe308fad0c95...

You can select models on Perplexity. He used Sonnet 4.5 Thinking

This may be interesting to you

"Vibe code is legacy code (val.town)" https://news.ycombinator.com/item?id=44739556

Fortunately, the GitHub README opens with a list of links to install ublock Origin.

https://github.com/gorhill/uBlock?tab=readme-ov-file#ublock-...

If you are a heavy user, you should use pay-as-you go pricing

if you’re a heavy user you should pay for a monthly subscription for Claude Code which is significantly cheaper than API costs.

Humans are far less deterministic than LLMs, yet presumably are acceptable for writing program source code?

Those “weird invocations” are called English.

The reason I initially got interested in Claude was because they were the first to offer a 200K token context window. That was massive in 2023. However, they didn't keep up once Gemini offered a 1M token window last year.

I'm glad to see an attempt to return to having a competitive context window.

See the authors response. He or she says it doesn’t matter either way

https://news.ycombinator.com/item?id=44706580

You should know that you are supposed to open the CLI (Claude Code, Gemini, ...) in your project directory and only use it to modify files within your project directory. This is meant to protect from problems like this.

Your "straightforward instruction": "ok great, first of all let's rename the folder you are in to call it 'AI CLI experiments' and move all the existing files within this folder to 'anuraag_xyz project'" clearly violates this intended barrier.

However, it does seem that Gemini pays less attention to security than Claude Code. For example, Gemini will happily open in my root directory. Claude Code will always prompt "Do you trust this directory? ..." when opening a new folder.

This is clever! and potentially useful too.

Have you done any testing to determine how precise and accurate this is? I suspect their must be a lot of variance between laptops, since this isn’t an intended use case.

This is a cool proof-of-concept! It reminds me of https://oasis-model.github.io/. which friends and I had a lot of fun with

The first link was the textbook I used for my computer architecture course last semester and I concur. This was the first time our professor taught ARM instead of x86_64 because he believes ARM is the future.

There are plenty of LLM use cases where the output isn’t meant to be read by a human at all. e.g:

parsing unstructured text into structured formats like JSON

translating between natural or programming languages

serving as a reasoning step in agentic systems

So even if it’s “too fast to read,” that speed can still be useful

This is cool. I think faster models can unlock entirely new usage paradigms, like how faster search enables incremental search.

I’m sure you could build a prototype add on to do this pretty quickly with Claude Code or the like

I think you misunderstood him. Yes, it’s possible to CREATE a useful webpage without JavaScript, but many EXISTING webpages rely on JavaScript to be functional.