👋 Hi, I’m Drew. I mainly post insightful or humorous quotes and comments I find elsewhere on the internet. 🔗 www.drewrobson.consulting

”Claude gave me the wrong answer, or it lied. Of course, it didn’t know it was lying because it’s not a human being; it’s a large language model... It had skimmed and sampled a lot of television reviews, but they never give away the ending in online criticism, so it was misled by the things it had read about the show.”

Margaret Atwood says the problem with AI is ‘garbage in, garbage out’ | The Verge

One of the more annoying problems with the current tech industry is that there are actual problems that these people could solve, many of them created by their own products. Meta has repeatedly made Facebook and Instagram worse to increase engagement, something it could easily reverse if not for the fact that it might lose Zuckerberg a dollar. Microsoft Word, Teams and basically every part of the Office suite is a bloated piece of shit, and based on the many conversations I have with people at the company, there’s plenty of talent that could fix them in seconds. Google Search blows like working in a fan factory because Sundar Pichai and Prabhakar Raghavan made results worse to increase the amount of searches people make, and Google Home is the ugliest and most-confusing app in history — both things that could easily be fixed if Pichai gave a shit about anything other than growth.

Premium: Notes From The Bubble, Volume 1

What makes for a fun job? Opinions vary, but I think it’s agency and impact. Agency means that your colleagues trust you to individually do your best work. It means not being told what to do, nor when to do it. It means working when you want to, and taking your kid to the doctor when you need to. And it means freely discussing designs with colleagues, with mutual respect and without pressure or politics. Impact means that the work you do directly improves the lives of our customers, and therefore makes the company better.

For example, we started out doing Scrum. But the sprint cadence made us take a single hard “do we release or not?” decision every two weeks, and it made no sense. Why not just ship a feature when it’s done? So we ditched the idea of sprints and switched to Kanban. Then we noticed that our Trello kanban board went outdated all the time, but it stopped nobody from shipping. So we cut that out too. Daily standups were interrupting half the team’s flow, so we cut those out too. We kept cutting things out, and by now, we’re a company with no deadlines, no obligatory meetings, no fixed working hours, and no bosses giving you tasks.

This works because, and only because, the entire company consists of people who move the fastest when they’re not held back by process. People who don’t need to be told what’s the most important thing to do right now (and who know to ask if they’re not sure). People who prefer to dive deep and emerge victorious, people who know how little sense it makes to try to split a performance problem into estimated “story point” tasks on sticky notes. People who know how to scope work down into small chunks, so we ship fast increments whenever we can.

But in the end, a pretty photo on the “careers” page touches you once. The freedom and trust to do your best work every day touches you all the time.

A no-nonsense working culture

I’ve stopped trying to debate software developers on LLMs. You might have noticed if you’ve been following this blog. It’s a fruitless debate. Even if the believers in agents and copilots could be budged on empirical grounds, and the past few years have given us plenty of evidence that they can’t, this is still a crowd that is explicitly fine with using tools that are themselves deeply unethical.

Debating people who look past “chatbot psychosis”, the dismantling of the education system, the gendered abuse, the generated CSAM images, the overt attacks on the media industries, or the ultra-right’s glee about “AI”, by showing them a well-constructed academic study is never going to work.

'AI' is a dick move, redux

Even the most staunch capitalist out there must acknowledge that this isn't it. There's no competition. There's no innovation. There's only a bunch of dumbasses who call themselves "product managers" whose whole job is to figure out novel ways to fuck you over, lock you in, and leave you with very little recourse. It's no longer about making cool shit that people like, if it was ever that. It's about setting the pieces in a way that you simply have no other choice. The highest achievement for this kind of enterprise today is not to have happy customers: is to get to the point where they can say "What are you going to do? Leave us for the competition? There's no competition."

Be interesting

This isn't about China — it's so much fucking easier if we let it be about China — it's about how the American tech industry is incurious, lazy, entitled, directionless and irresponsible. OpenAi and Anthropic are the antithesis of Silicon Valley. They are incumbents, public companies wearing startup suits, unwilling to take on real challenges, more focused on optics and marketing than they are on solving problems, even the problems that they themselves created with their large language models.

By making this "about China" we ignore the root of the problem — that the American tech industry is no longer interested in making good software that helps people.

Be interesting

When you come from an environment of cutthroat competition, of constant backstabbing, of extreme individualism where you have got to get yours, fuck everyone else, collaboration seems like a completely absurd idea. What's in it for me? It should also be clear by now that the whole "tech" scene these days is exactly like this. Get your VC money, sell out, fuck everybody else. If the game is not a zero-sum one, it's not worth playing: it's not sufficient for me to win. You need to lose.

Pour one out for community

So, instead of having a place where a lot of questions have been collectively answered satisfactorily, indexed, archived, and catalogued, we are choosing the method of individually asking the same questions over and over again. If I have the exact same question as you do, I need to ask the LLM again. All the tokens, and energy, and GPUs, and training data, and models, are doing the same work twice. Or thrice. Or a billion times. Instead of a thousand people finding one answer that's one search away, we are asking the same question a thousand times. Instead of a thousand people typing "How do I reference a specific list item in Python" on the search box and getting answers, we're asking LLMs to compute and generate the same answer a thousand times. And that's only possible for those who even have access to these tools.

Pour one out for community

Musk has spent his entire career using his money to convince people he’s smart. This is why Musk sued the actual founders of both the platform that would become Paypal and, also, Tesla to make sure he’d have to be called a founder of both. And why he has spent considerable energy over the last two decades trying to convince credulous journalists and idiotic politicians that he was irl Tony Stark. Now he’s found the dumbest, least inquisitive, most media illiterate people to ever walk the Earth as his new marks: Republicans. So he bought their favorite website and set up his super PAC and is now, once again, using his money as a way to convince American conservatives that, as Trump put it, “he knows those computers better than anybody.” Maybe Trump thinks X posts helped him win (they absolutely didn’t), maybe Trump thinks Musk actually rigged the election (also doubtful). But I imagine it’s similar to when you help your grandfather reset his Netflix password. Doesn’t matter what happened in the computer, but it worked. The PDF rotated. And if Musk did have any actual impact on last year’s election, it was almost certainly thanks to the $290 million he spent on Trump’s campaign, around $50 million of which were spent on million-dollar giveaways like the one he tried again this week. But, once again, did it matter? Seeing as how things played out in Wisconsin, likely not. Which is a problem because the only thing Musk is truly a genius at is burning through unimaginable amounts of money on dumb shit no one cares about.

I think Elon's getting Vivek'd after this one, folks