Skip Navigation

Mikina

@ Mikina @programming.dev

Posts
5
Comments
304
Joined
3 yr. ago

  • Vibe management

    Jump

  • Because their LLMs suck, and they still want their devs to be productive, while watching their usage to train the internal models.

    They give you a choice, and no one is using the internal models.

  • Microsoft Drops Claude Code After Budget Overrun

    Jump

  • I kind of respect that. I wonder what happened to him.

    Meta has some tolerable studios, and I can kind of get how it could happen to you that you end up working for Meta. I know a few people from Beat Games, for example, and they work on a great game, are an amazing team, that has basically not much to do with what Meta does.

    It's difficult to stand your ground and just be "No, I'm not workibg for facists", when the team you are in is full of amazing people burning Meta's money to work on their dream game. You could leave, but you're just leaving your friends behind that will loose a lot of domain knowledge and a good programmer, and it will not change anything. I also know a few contractors from 3rd party codevelopment studios that got sent to Meta's 1st party studios, and there it's even worse - they could say no, but then it's cutting off an income source from Meta that might as well be keeping the studio afloat.

    So it's not just black and white.

    But anyone working on core Meta social networks/services/ad platform/algorithms?

    Fuck those guys. There's no excuse.

  • Vibe management

    Jump

  • I've been kind of forced to use AI for my work, and since I had an unlimited access to whatever model I needed, I figured "why not try". I was able to find my workflow that works, use it to explain the architecture of the code surrounding my bugfix/feature (I work in gamedev, on a project I'm only a contractor for, so I don't know the codebase too well), make me a documentation, then draft a plan how to implement it. Implement it, then I just look through it for ideas, combine it with my domain knowlege to see if it missed any obvious things or solutions (which it usually does in a larger projects), and then build my own solution from scratch based on what I know, and what it suggested.

    The part where it explains the architecture and data/systems flow is invaluable and it does make ot faster than I could have parsed through unknown code, while being verifiable enough to be trustworthy. It's a good kickstarting process. Do I need it? Not really, but it would take me longer.

    But. It eats tokens like hell. My average monthly token usage is around 800m tokens.

    I've been told I'm not using the AI enough in my workflow, because I write my PR comments and don't use AI for code reviews.

    I'm seriously considering just leaving IT altogether. It's just eroding my reverse engineering / codebase orientation skill, while I'm replacing it with something that costs the same amount as my sallary for doing it slightly faster/more easier, and the price will only get a lot worse. I don't get it. How can't they see it?? How can they look at "Oh, he's costing us 2000$ a month in AI use [at current prices], but solves 3 instead of 2 tasks a week, while slowly loosing vital skills", and say "that's worth it! But he could use it more."

    I hate it. Fuck managment.

  • Kindle loyalists scramble as Amazon turns page on old e-readers

    Jump

  • The first thing I did with my kindle that I got 10 years ago was to turn on Airplane mode and never turned it off since. I can upload books there via usb just fine, and I never had an issue.

  • Microsoft Drops Claude Code After Budget Overrun

    Jump

  • I mean, I can imagine it's a very real expense.

    Not sure what is the difference between Meta and Microsoft in term of employee size, but if Meta is burning 300 000 000$ monthly on tokens (whichis a number reported by their internal tracking tool that leaked assuming 5$ per 1M tokens), then it's not cheap :D

    In a 30-day period, total employee usage on the dashboard exceeded 60 trillion tokens, and the highest-ranked individual user averaged 281 billion tokens. Using the least expensive version of Claude Opus 4.6, which costs $5 for every million tokens, that one user alone could have cost Meta more than $1.4 million.

    ⁨https://fortune.com/2026/04/09/meta-killed-employee-ai-token-dashboard/%E2%81%A9

  • Vibe management

    Jump

  • Meta was spending ~300 MILLION dollars per MONTH on tokens around a month ago.

    I know they are using Claude, so the price estimation of 5$ per 1M tokens is kind of true, more or less.

    Tell me how is this sustainable in the long term?

    In a 30-day period, total employee usage on the dashboard exceeded 60 trillion tokens, and the highest-ranked individual user averaged 281 billion tokens. Using the least expensive version of Claude Opus 4.6, which costs $5 for every million tokens, that one user alone could have cost Meta more than $1.4 million.

    ⁨https://fortune.com/2026/04/09/meta-killed-employee-ai-token-dashboard/%E2%81%A9

  • More than 170 arrests in Met's Croydon facial recognition trial

    Jump

  • Of course it will lead to reduced crime, and allow them to solve more of them. That's kind of expected, when you can find and track where is anyone 24/7.

    But then you inevitably get a far-right governemt, because that's the rapidly rising trend all over the world, and suddenly 'crime' also means being trans, or feeding homeless, or not being far-right enough, or speaking up about how maybe we shouldn't jail trans people.

    10+ years ago, facial regonition like this would still be controversial, but you could maybe kind of have a benefit of the doubt that it might be worth the extremely reduced crime rate.

    Now, when democracies are falling left and right and it's only a matter of time when your country starts their own version of concentration camps, it's a really stupid fucking idea.

    So, fuck this bullshit propaganda piece.

  • Fears grow that age verification coming to VPNs as a British research firm labels them a 'loophole'

    Jump

  • A reminder that Tor's Snowflake exists, and running a node can help and is as easy as just installing a browser extension and never thinking about it again.

    https://snowflake.torproject.org/

  • If development were an RPG game, which RPG class would you be?

    Jump

  • If you're interested in playing corporation-based, slightly occult, TTRPG, there is Corp Borg that fits the bill.

    https://heltung-storytelling.itch.io/corp-borg

    Corp Borg is an award-nominated (NDU, The Indie Groundbreakers), rules-light, office-crawl OSR tabletop RPG. While inspired heavily by MÖRK BORG aesthetics and gameplay, it's a standalone game - all you need to play is this book.

    The world of Corp Borg is bleak, bloody, and brutal, exactly like corporations are. They rule the world of Corp Borg, as they do in real life, but are a bit more obvious about summoning demons, making blood sacrifices, and performing occult rituals.

  • Bluesky's new AI tool Attie is already the most blocked account other than J.D. Vance

    Jump

  • Good that it's decentralized and people can just move on to another provider, since this one is doing this, haha, right?

  • Jetbrains is using AI-generated splash screen art

    Jump

  • I dropped my subscribtion the moment the AI bullshit started, and keep using my perpetual/fallback license for the version before that.

    Once you pay for a year, you can keep using that years version indefinitely, save money and as a bonus you don't get more AI shit stuffed in your throat.

  • With Reddit flirting with requiring Age verification, the next Rexit might be around the corner, are we ready?

    Jump

  • I don't want people from Reddit here.

    The fact that half of Twatter moved to Bluesky instead of Mastodon is a blessing.

    ActivityPub is by design a data harvesting goldmine, the fact that it flies under the radar is the only saving grace.

  • Thoughts on slowing the fuck down

    Jump

  • You can do a task pretty well if you nudge the AI, have it write an exact explanation about every part of the architecture, code and data flow it's working with and throw relevant files into context, and correct anything that's wrong before you send it to do the task. You still have to review, but I didn't have to correct much in my experience.

    But that burns like 20$ of tokens per task, at current prices that are way below the costs AI companies are paying.

    While it does help me, especially with parts of the codebase I'm not familliar with, it's not sustainable, and it's actively and very quickly robbing me of my skills and knowledge. It's really a bad idea to use it, in two years time you'll be royally fucked once they raise prices to recover the trillions they are loosing right now.

    So, however tempting, I simply don't use it. I won't throw away years of college and experience just to do a task a little bit faster today.

  • Reckless, a chess engine written in Rust, has made it to the final of TCEC season 29 and will battle Stockfish for first place

    Jump

  • What does time control of 120 minutes mean? That they have 120 minutes per turn?

  • Indie devs are the true heroes of OSS

    Jump

  • Battlefield has parts that run on Godot.

    I don't see EA on the donors page.

  • The ‘European’ Jolla Phone Is an Anti-Big-Tech Smartphone

    Jump

  • Are there even any European chipmakers? I guess there are, but never really looked into it.

  • Big brother wants to help

    Jump

  • I don't think that"s the reason for the absird lobbying, which is why it's getting through.

    Governments aren't that competent to get this through at scale, but corporations can and have a very good reasob to do so.

    I believe that it's a lot more probable that the reason is profit - social networks have a problem with bots. Advertisers don't want to pay if 90% of impressions are bots, and you can't really solve that problem easily.

    Age verification by ID solves this, and if they even can lobby hard enough to outsource the costs to mandatory OS level veryfication, so they just call one OS api, it's even better.

  • Surge in Systemd forks after the latest changes

    Jump

  • They should make the API call for apps to query that value a per-system/boot randomly generated signature, so it's impossible to use while also complying with the law.

  • Surge in Systemd forks after the latest changes

    Jump

  • I'm mostly interested in how will they handle giving the info to apps. If it'd let me to block or fake the request depending on what I currently need (just prompt me every time an app asks, and let me choose the bracket), I'm good.

    Tbh, most sites that are slowly getting targeted by age verification laws are things I'm kind of addicted to and have been trying to drop for a long time. A "scan your face or id" dialog would be a good reminder to finally cold turkey it. It's one of the things I hate more than however much I need their platforms.