yojo

I read the balls as “houses”, though the phallic spire emerging from them dead-center is also very on-brand

Oh, and the memory use! I run a lot of concurrent sessions. 3 gigs for a terminal window is ludicrous.

I have been working all day every day in Claude. I loathe their bug-ridden UI. Every release is a new crop of bugs, sometimes the old ones get fixed, usually not.

Any kind of scrolling back, copying text, using their menu system - basically anything that isn’t typing characters has had/still has unaddressed bugs.

OpenAI shipped a competitive model and I’m over in Codex now. I have yet to hit a bug.

If you’re holding the SOTA crown, people will put up with your buggy mess. As soon as that crown slips your pile of trash becomes a huge liability.

I’ve experienced this in Claude Code - it’ll start asking me if we can submit the PR at the end of every turn.

Google at least is serving AI results on SRPs billions of times a day, and has pre-existing expertise in data center buildouts and custom silicon.

They have one of the more compelling cases for rolling their own.

Schumpeter is one of their columns; this is an editorial piece, not one of their news articles.

The regular stuff is still well written, IMO.

Meta is straight evil. It undermines the institutions of democracy and it negatively impacts its users mental health, all in service of selling your data to advertisers so they can better goad unnecessary consumption.

If I learn you work at Meta, I will judge you as at best lacking a moral compass and treat you appropriately.

Apple has problems, but is a lot closer to morally neutral. Ditto for Netflix.

Amazon has hollowed out local retail/is also bad for society, though not on Meta’s scale. But you sell your soul more cheaply there.

Internally, it’s important to understand that every ask should have a business goal associated with it. The thing being asked for is rarely (never?) the only way to accomplish that goal.

Great engineers focus on the customer or business need and find/propose alternatives that are possible.

I have worked at places that have negotiated flat percentage discounts on all AWS spend.

This explanation seems plausible to me.

You can hook traditional SAST into your coding tool, and get cheap-ish realtime detection for some classes of vulns while coding.

You can optionally layer LLM diff scanning if you want to burn some tokens on your tokens. Modern tools can catch some impressively subtle issues.

Before unlimited PTO took off, standard in FAANG-like (US) industry was ~12 corp holidays + 15-25 days annual PTO, depending on seniority.

I don’t know anyone who takes the high end of that anymore, especially senior/staff folks.

Unlimited PTO takes a liability off the company’s books, and makes every time off request a negotiation.

Two things:

1) Unlimited PTO is a scam. Ask anyone who ever got paid out six weeks salary when they changed jobs.

2) “the craft” is doing some heavy lifting here. I happen to enjoy AI-assisted dev, but it is nothing like the work that drew me to the industry.

Otherwise agreed on all counts.

I’m taking it as a given that any sufficiently large social network is a gigantic propaganda machine of interest to domestic and foreign nation-state actors.

Entertaining the thought experiment where all the normies join the fediverse: now you’ve got a big juicy target maintained by hobbyists.

When it’s Lazarus Group vs Randall, the over-worked sys admin who stood up a node in his spare time, who do you think wins?

Social networks are cancer. Just ban the lot of them and move on.

Me? No. My kids? I think they already have. I don’t allow YouTube in our house, but they for sure watch slop with friends.

Oregon decriminalized drugs in 2020 and the experiment is widely viewed as a failure by both sides of the political spectrum. The Democratic legislature rolled it back four years later.

It doesn’t necessarily follow that it’s impossible to have a legalized or decriminalized regime that works, but it is non-trivial to get right.

You should probably be more worried about the flame retardants in the rebond carpet pad. Better these days, but older installed stuff had non-trivial PPM.

DropMix was an NFC DJing game that used RFID playing cards. It’s been done/could be done again.

Meta’s bottom line is driven entirely by their ability to uniquely and persistently identify users for the sake of advertising.

Anything that makes it harder for a user to escape their dragnet is a win.

I got a Zune on clearance when they were shutting down the whole mess (a nice fugly brown one for 80% off).

Build quality was rock solid, UI felt premium, and it mostly just got out of my way/let me play tracks. They were great little media players, as good as or ahead of the equivalent iPod.

Hardware was really good out of MSFT at the time, although when mistakes were made (e.g. the RROD Xbox 360 debacle), the broader organization seemed allergic to making thing right.

You might try adding this branch protection rule to require conversation resolution: https://docs.github.com/en/repositories/configuring-branches...

I also follow this approach. I just flip the flag on the PR I’m reviewing to off before submitting my approval.

We also have most of our repos set to block if unresolved comments. I think it’s a flag on branch protection rules

option + hyphen gives you an en-dash (–), which is easier to type and I am guilty of way overusing/misusing.

My work has a “donuts” slack channel for this. You find an unlocked computer you post “donuts on me!” Social pressure says they buy the office donuts.

Still get a few a week, but at least it’s public and amusing.

I caught myself saying “you’re absolutely right” to my wife last night, unironically. This was 100% not in my vocabulary six months ago.

If I spend 40 hours a week talking to anybody, some of their language or mannerisms are going to rub off on me. I can’t think of a compelling reason why a human-sounding chat bot would be any different.

+1

I’ve been driving Claude as my primary coding interface the last three months at my job. Other than a different domain, I feel like I could have written this exact article.

The project I’m on started as a vibe-coded prototype that quickly got promoted to a production service we sell.

I’ve had to build the mental model after the fact, while refactoring and ripping out large chunks of nonsense or dead code.

But the product wouldn’t exist without that quick and dirty prototype, and I can use Claude as a goddamned chainsaw to clean up.

On Friday, I finally added a type checker pre-commit hook and fixed the 90 existing errors (properly, no type ignores) in ~2 hours. I tried full-agentic first, and it failed miserably, then I went through error by error with Claude, we tightened up some exiting types, fixed some clunky abstractions, and got a nice, clean result.

AI-assisted coding is amazing, but IMO for production code there’s no substitute for human review and guidance.

The TL;DR dos not seem to match the rest of the article.

They claim the agents reliably generated a week’s worth of dev work for $20 in tokens, then go on to list all the failure modes and debugging they had to do to get it to work, and conclude with “Agents are not ready to autonomously ship every integration end-to-end.”

Generally a good write up that matches my experience (experts can make systems that can guide agents to do useful work, with review), but the first section is pretty misleading.

Having spent a couple years rehabbing a 100 year old house, I’m convinced the trades will be the last thing to go. When the building you’re working on has been ship-of-Theseus’d by 3 generations of home owners, everything is out of distribution.

When a robot can reliably do this work, I think it can reliably do any human job that requires physical ability and judgement.

LLMs rarely if ever proactively identify cleanup refactors that reduce the complexity of a codebase. They do, however, still happily duplicate logic or large blocks of markup, defer imports rather than fixing dependency cycles, introduce new abstractions for minimal logic, and freely accumulate a plethora of little papercuts and speed bumps.

These same LLMs will then get lost in the intricacies of the maze they created on subsequent tasks, until they are unable to make forward progress without introducing regressions.

You can at this point ask the LLM to rewrite the rat’s nest, and it will likely produce new code that is slightly less horrible but introduces its own crop of new bugs.

All of this is avoidable, if you take the wheel and steer the thing a little. But all the evidence I’ve seen is that it’s not ready for full automation, unless your user base has a high tolerance for bugs.

I understand Anthropic builds Claude Code without looking at the code. And I encounter new bugs, some of them quite obvious and bad, every single day. A Claude process starts at 200MB of RAM and grows from there, for a CLI tool that is just a bundle of file tools glued to a wrapper around an API!

I think they have a rats nest over there, but they’re the only game in town so I have to live with this nonsense.

- This is partly b/c it is good at things I'm not good at (e.g. front end design)

Everyone thinks LLMs are good at the things they are bad at. In many cases they are still just giving “plausible” code that you don’t have the experience to accurately judge.

I have a lot of frontend app dev experience. Even modern tools (Claude w/Opus 4.6 and a decent Claude.md) will slip in unmaintainable slop in frontend changes. I catch cases multiple times a day in code review.

Not contradicting your broader point. Indeed, I think if you’ve spent years working on any topic, you quickly realize Claude needs human guidance for production quality code in that domain.

There are a few modern artists who mix cool techniques to great results and get recognized. Olafur Eliasson and James Turrell come to mind.

Damien Hirst is a more polarizing third contender.

Edit: also Yayoi Kusama