hooch

Since the last time I tried, Apple Maps has come down from plotting some 30 petrol stations in the Sydney CBD to 2. There are in fact none.

Apart from needing to copy and paste destination addresses from Google/Google Maps every time I use Apple Maps, it's great.

True, it first appeared around 2011

Sounds like you ought to try Claude Desktop.

Stripe is a wrapper too. Single API key and unified DX atop a world of divergent providers.

Also here, it's a really cool conversation https://syntax.fm/show/1029/the-workflow-of-the-future-with-...

Making you pay ~12% fee on foreign country taxes collected by eBay, when those taxes can’t be gamed like postage was, is a bridge too far.

I used to offer listings to the world, now I just add Japan for the odd rare item.

This

No 5.5 Pro

confirmed

For some reason the latest Claude Desktop release from Anthropic threw off its Claude branding and charm to chase after bland Codex Desktop app look and feels.

Maybe they think OpenAI is doing something right?

Last time I looked Copilot's context window for Anthropic models was something like 150,000 tokens only.

it's as if the icons get added by a lazy LLM prompt during CI

Was doge ever anything more than a "get root, grab the data, and run" operation?

Sequoia is the new Snow Leopard

Changed into a directory recently and fired up the qwen code CLI and gave it two prompts: "so what's this then?" - to which it had a good summary across stack and product, and then "think you can find something todo in the TODO?" - and while I was busy in Claude Code on another project, it neatly finished three HTML & CSS tasks - that I had been procrastinating on for weeks.

This was a qwen3-coder-next 35B model on M4 Max with 64GB which seems to be 51GB size according to ollama. Have not yet tried the variants from the TFA.

The artist https://www.carlburton.io/

Too right. Drilling into the domain from first principles and with critical faculties enabled unlocks so much more value, because the engineer can then see much better ways to solve problems.

It's just how the memory controllers work in Apple Silicon's unified memory architecture.

You also get much higher memory bandwidth as you move up the scale, from 153GB/s at the bottom end to 546GB/s at the top end.

But Paramount only recently signed a massive deal for Southpark, which is outrageous in its handling of Trump.

M3 support still over yonder, unfortunately.

Actually the internet was not down at all. It was perfectly up.

tldr;

Before:

    let process = Process()
    process.executableURL = URL(fileURLWithPath: "/bin/ls")

    let pipe = Pipe()
    process.standardOutput = pipe

    try! process.run()
    process.waitUntilExit()
    let data = pipe.fileHandleForReading.readDataToEndOfFile()
    if let output = String(data: data, encoding: .utf8) {
        print(output)
    }

After:

    let result = try await run(
        .name("ls"),
        arguments: ["-1"],
        output: .string(limit: 1 << 20)
    )
    print(result.standardOutput ?? "")

Always been one more iteration of 3nm in the plan

From a companion article: "For instance, an F-35 contains more than 900 pounds of REEs, an Arleigh Burke–class DDG-51 destroyer requires approximately 5,200 pounds, and a Virginia-class submarine uses about 9,200 pounds."

It was still shown in Israel though and can be found online.

The cadence is not just about shipping marketable products and features.

It's also about iterating on internal capabilities, processes, technologies across design, sourcing, manufacturing, etc.

What makes it into each year's phones is what's had the kinks ironed out, and all the commercials sorted.

Uber also had the good sense to not bake a narrow service category into its brand.

Quite exceptional