jcheng

Can you say more? Why isn't it neutral or slightly positive? I would assume that a KYC provider would want to protect their reputation more than the average company. If I were choosing a KYC provider I would definitely want to choose the one that had not been subject to any privacy scandals, and there are no network effects or monopoly power to protect them.

That’s what this is. It’s just defining two types of subagent relationship (spawned and forked) and providing the minimal MCP API for controlling them. It’s up to the LLMs when and how to use subagents.

The revenue curves do indeed look a bit different:

https://www.macrotrends.net/stocks/stock-comparison?s=revenu...

Even more so if you compare EBITDA:

https://www.macrotrends.net/stocks/stock-comparison?s=ebitda...

Would’ve been perfectly readable and no larger if they had used newline instead of pipe.

Yeah... and besides managerial skills, also product (using the word loosely) sense, user empathy, clarity of vision, communication skills. They've always been multipliers for programmers, even more so in this moment.

> 95% of Claude and 5% of you, while still better than me (and your average Joe), is nowhere near the same jump from 95% Claude and 5% me.

I see what you're saying, but I'm not sure it is true. Take simonw and tymscar, put them each in charge of a team of 19 engineers (of identical capabilities). Is the result "nowhere near the same jump" as simonw vs. tymscar alone? I think it's potentially a much bigger jump, if there are differences in who has better ideas and not just who can code the fastest.

> For others, LLMs remove the core part of what makes programming fun for them.

Anecdotally, I’ve had a few coworkers go from putting themselves firmly in this category to saying “this is the most fun I’ve ever had in my career” in the last two months. The recent improvement in models and coding agents (Claude Code with Opus 4.5 in our case) is changing a lot of minds.

An engineer on my team who is working on TUI stuff said that avoiding the flicker is difficult without affecting the ability to copy/paste using the mouse (something to do with "alternate screen mode"). I haven't used OpenCode (yet) but Google does turn up some questions (and suggested workarounds) around copy/paste.

Arguably since 3.5, at least for coding and tool calling

To make this right they’d just have to amend the first part to “secondary transaction with shareholders at some price that implies a 20b valuation”.

Has there been any evidence yet that the VCs got paid for their shares but the left behind employees didn’t?

And text figures! And proper small caps!!

That’s exactly what it does, I’ve found it completely un-confuses Claude Sonnet 4.5.

That makes so much sense, it would make a great MCP. Maybe something similar for DOM manipulation; extracting text out of big, noisy HTML pages using a combination of Find Text with selector return values, and a DSL for picking and filtering DOM trees.

I take these early reports (less than a week or two after a major model release) with a grain of salt. It takes time to get to know a model, and maybe there's some selection bias in who's posting within 1-2 days of getting access.

(Although in this particular case, the very different knowledge cutoff makes it a lot easier to believe)

Personally, CLAUDE_BASH_MAINTAIN_PROJECT_WORKING_DIR=1 made all my cd problems go away (which were only really in cmake-based projects to begin with).

Seems like an opportunity for someone

They met tonight! This is so insane!

Reminds me of Stephen Colbert's roast of George W. Bush at the 2006 White House Correspondents' Dinner:

> The greatest thing about this man is he's steady. You know where he stands. He believes the same thing Wednesday that he believed on Monday, no matter what happened Tuesday. Events can change; this man's beliefs never will.

At the time, Java. J2EE (entity beans and session beans), Java Server Pages, Apache Struts. I think it's hard for people who didn't live through it to appreciate just how painful it was to work in that stack circa 1999-2003, like it just doesn't seem plausible that the whole industry would go along with a stack with that much boilerplate for that little added value.

Back then, Ruby and Rails opened a lot of people's minds to the idea that we were allowed to make "delightful" a consideration in API design, not just S.O.L.I.D. or whatever. These days, there are way more mainstream languages, frameworks, and communities that take humane APIs seriously.

Why do you think this is worse than someone saying about Java: "some coworker put `this.` on a variable relating to sessions and suddenly people started seeing each other's accounts"? Because it's less obvious what "@" means rather than "this."?

Edit: I forgot that Java has implicit `this`! That's so, so much worse!

Generative, not general

Interesting that they run them upside down.

Automatic speech recognition and speech to text models are also growing up real fast.

Wow, it's true--Terminal is <canvas>, while the editor is DOM elements (for now). I'm impressed that I use both every day and never noticed any difference.

Sure, I’m with you on your larger point.

Do you have a citation for that factoid?

Seems almost quaint in late 2025 to object to a workable technique because it "seemed like a bit of a hack"!

AFAIK, every tool out there that lets you do oauth with your Claude Max plan is doing so with the same copy-pasted client id/secret that are extracted from Claude Code. It's not clear at all that this is above board, and when I asked our Anthropic rep, they encouraged us not to do this.

If there's an official way for 3rd party devs to piggyback on Anthropic plans, someone please tell me how!

Slightly off topic, but I wonder if Quarto would work for you (https://quarto.org, disclaimer, I work for the company that maintains it). You write Markdown files with code blocks, and they're rendered into nice looking documents in a surprising range of possible form factors (see https://quarto.org/docs/gallery/).

You can also use .ipynb files as input instead of Markdown files and output to those same formats.

Traditionally, Quarto executes Python code at render time, but there's an extension called Quarto Live that lets you also incorporate Pyodide code blocks for the reader to use. https://r-wasm.github.io/quarto-live/ (I have not used Quarto Live before and have no idea what the limitations are.)

512MB file of incredibly compressible data, then?