Forget-Me-Todd (@Kelsarrin) header
Forget-Me-Todd (@Kelsarrin) profile picture

Булочка без корицы

Revachol
Joined October 2011
"Rockstar меня уволили в эту среду" - и забыл проверить фальшивое письмо из 2025 года Очередной ИИ бро-хайпожор
Rockstar Games fired me on Wednesday. GTA 6 launched on Thursday. I was a Senior Engine Engineer at Rockstar North, working on RAGE infrastructure and internal tooling. I've been with the studio for six years. I shipped RDR2 on PC. I was part of the core team for GTA 6 since pre-production. What I'm about to tell you is not about me getting fired. It's about something that every game studio, every tech company, and every software corporation on earth is about to face and none of them are ready for it. Their entire business model is about to collapse. And I have the proof because I accidentally did it on a Saturday afternoon. In August I started experimenting with AI agent workflows. Not on company time. Not on company hardware. On my personal laptop, on a weekend. I wanted to see what would happen if I described one of our internal tools to an AI the way I would explain it to a new hire. I opened Claude Code running Fable 5.2. I described our mission scripting pipeline. Not from documentation. Not from source code. From memory. The way any senior engineer talks about systems they built - the architecture, the data flow, the edge cases, the design decisions. The AI asked me follow-up questions. I answered them the way I'd answer a junior developer during onboarding. Normal engineering conversation. The kind that happens in every standup, every code review, every lunch break at every tech company in the world. Then it started building. Eleven hours later I had a functionally complete recreation of a system that took our team fourteen months to ship. I need you to understand what "functionally complete" means. This is not a rough sketch. This is not a prototype. The state machine was correct. The event pipeline worked. The edge cases were handled. The build compiled. It ran. I tested it against real scenarios from memory. It passed every single one. Fourteen months of work by a team of engineers. Rebuilt from a conversation. In eleven hours. By one person who wasn't even trying that hard. But that's not the part that should scare you. Here's the part that should scare you. I didn't do anything that any engineer at any company couldn't do right now. I didn't hack into Rockstar's servers. I didn't export a single file. I didn't copy a single line of source code. I didn't use any company credentials. I didn't access any internal repository. I didn't violate any access controls. I didn't even open my work laptop. I talked to an AI about how software works. That's it. That's the whole crime. Every engineer reading this has done a version of what I did. Every time you've described your company's architecture to ChatGPT to debug a problem. Every time you've explained a system design to Claude to get a second opinion. Every time you've pasted a stack trace and described the surrounding code to Copilot. You've been doing what I did. The only difference is you didn't compile the output. I compiled it. And it worked. And that's when everything went wrong. I brought the prototype to my technical lead on Monday. I was excited. I thought I had discovered something that could save the studio thousands of engineering hours. I thought this was the future of how we build tools. I thought they'd promote me. He went quiet. He asked me to send him the files. I did. 48 hours later I was in a room with Legal, HR, and my department head. No technical leads in the room. No engineers. No one who would understand what I actually did. Just lawyers. They told me I had "reproduced proprietary systems and internal tools outside of the approved development environment using unauthorized AI agent workflows." They told me the resulting build was "functionally complete" and represented a "material breach of our Security and Intellectual Property policies." Then they told me the number. $47,500,000. Forty-seven point five million dollars in estimated damages. For a conversation I had with an AI on my personal laptop on a Saturday. They presented me with a termination letter. Effective immediately. They had already revoked my access across every Rockstar system before I walked into that room. Badge. Email. VPN. Jira. Confluence. Perforce. Six years - gone before I sat down in the chair. The letter says they've referred the matter to external counsel. It says they're cooperating with federal authorities. It says I need to return all company hardware and removable media by Monday. I want to be very clear about what's actually happening here because the framing in that letter is designed to make me look like a spy. Like I broke into a vault and walked out with the blueprints. Here's what actually happened: I described how software works to a machine that understands software. And the machine built it. That's it. That's the $47.5 million crime. And here's why Rockstar is terrified. Here's why they went straight to lawyers instead of engineers. Here's why the number is $47.5 million instead of a written warning. Because if what I did works - and it does work, I have the build to prove it - then the fundamental premise of the entire software industry is broken. Every company in tech operates on one core assumption: our code is our moat. Our proprietary systems are our competitive advantage. It took us years to build this, and that time investment is what protects us. I just proved that assumption is wrong. The moat is not the code. The code can be rebuilt in eleven hours. The moat is not the architecture. The architecture lives in the heads of the engineers. The moat is not the years of investment. The years of investment can be compressed into a single conversation. The only moat left is the people. And the people can talk. Not to competitors. Not to journalists. Not to foreign governments. To AI. To a tool that anyone can download and run on a laptop. Rockstar's $47.5 million is not about punishing me. It's about sending a message to every engineer in that building: do not talk to AI about what you know. Do not describe your work. Do not explain your systems. Because if you do, those systems can be rebuilt overnight. And we will end you. That's not a security policy. That's a gag order on human knowledge. Here's the part that keeps me up at night. I did this by accident. I wasn't trying to steal anything. I wasn't trying to prove a point. I was running a weekend experiment. What happens when someone actually tries? What happens when a competitor hires three senior engineers from a game studio and sits them down with an AI for a week? What happens when every startup in the world realizes they don't need to build from scratch - they just need to hire people who've built before? This is already happening. It's happening right now. And the only reason you haven't heard about it is because everyone who discovers it gets a termination letter and a threat before they can tell anyone. I'm telling you. This is the termination email. I've redacted my name and personal details. $47.5 million. For talking to Claude.
Readers added context they thought people might want to know
Anthropic released Claude Fable 5.2 on 9 November 2026. The email shown is dated March 7, 2025 yet names Fable 5.2, and the post says the model was used in August, before it existed. tbreak.com/claude-fable-5… 9to5mac.com/2026/11/09/ant…
3
97
Какой шарик стал невыносимо скучный, господи помилуй
4
7
150
ребят, если кто-то сможет помочь нам с Батонове оплатить консультацию у нейрохирурга и хотя бы частично погасить долг в Витавет (там 45к), мы будем очень благодарны 🥺🙏 5536914169063991 тинёк 2202206393055004 сбер 2200153971889168 альфа 9219611572 тлф для сбп
ну, штош. у Батонове по мрт новообразование в лобной доле мозга, очаг 5мм. отсюда, собственно, и судороги, шатучесть и все остальное. на завтра записались к нейрохирургу в клинику сложного случая, будем смотреть, что с этой гажей можно сделать.
2
69
5
150
2,302
Небольшой тред советских мультфильмов, которые занимаются отдельный уголок в моем сердце. Начнём с невероятно трогательного "Тигрёнка на подсолнухе " Очень славная и реалистичная рисовка+ потрясающий голос Евгения Леонова
23
216
13
3,164
40,117
I hate the degradation of the word Luddite. They weren’t idiots afraid of technology. They were a labor movement concerned about poor working conditions, unfair pay, child labor. They broke machines cause that was one of the only ways they could put pressure on their employers
David Fincher Defends A.I. As ‘Cliff Booth’ Faces Generative A.I. Controversy: “I'm Not A Luddite. I Don’t Fear The Unknown” theplaylist.net/david-finche…
32
2,053
43
10,582
130,970
Forget-Me-Todd retweeted
People attacking Léa when Daniel Craig said the same exact thing… #noticing.
Léa Seydoux when asked who she thinks should be the next Bond Girl: “I don’t really care.” (variety.com/2026/film/global…)
15
190
2
3,815
30,222
Forget-Me-Todd retweeted
eva green in casino royale (2006)
46
718
70
16,553
240,195
Фух, у малого нет лейкоза! Одной проблемой меньше
1
16
320
Forget-Me-Todd retweeted
"If I were Sauron I would be good" is as good an encapsulation as any of the ethos driving Anthropic
Tech people understand a book challenge impossible
6
33
1
181
2,719
Мне наоборот кажется, что это круто. Мультипликаторы пробовали разные стили и направления, придумывали подходы у анимации без оглядки на "это популярно, это нет". Интересно сравнивать старые мультфильмы с нынешними 3д проектами. Мне бы хотелось, чтобы "страшного" было больше.
сегодня в тт узнала, что есть ссср версия мумитроллей, и я который раз убеждаюсь, что советские мульты уродские и страшные до нельзя, даже с низким бюджетом на анимацию я не понимаю, почему у художников настолько уродские образы в голове возникают для ЦА дошколят
1
25
554
Forget-Me-Todd retweeted
when u have a soulmate
Léa Seydoux when asked who she thinks should be the next Bond Girl: “I don’t really care.” (variety.com/2026/film/global…)
14
2,229
7
80,699
706,343
Я была очень застенчивым ребенком, не могла постоять за себя. Как-то зимой у нас уроки (вторая смена) закончились раньше, и я одна сидела в гардеробной, ждала родителей. Местная техничка ныла, что я мозолю ей глаза, потом начала на меня орать. Я вышла во двор, чтобы она от меня->
Какой самый добрый поступок по отношению к вам совершил незнакомец, которого вы больше никогда не видели?
4
3
2
1,823
83,690
отстала и стояла там на морозе. Мимо проходил старшеклассник, спросил, что я тут делаю одна, когда уже темно. Я рассказала. Он взял меня за руку и отвёл обратно. И сидел со мной, пока не приехали родители. Мне было семь, и я была глупа, а парню большое спасибо.
2
1
1,714
19,230
Тут ещё важно добавить, что школа, хоть и в центре города, стояла в тихом месте рядом с оврагом и рекой, и по вечерам там бывало довольно стрёмно. Через несколько лет на мою одноклассницу в школьном дворе напал ублюдок с ножом, когда она пришла забирать с уроков младшего брата
1
1,263
17,950
Forget-Me-Todd retweeted
Что я узнала о ГТА 6 пока что: Христианская Не христианская Воук Расистская Феминистская Антирасистская Там толстые женщины (кто-то впервые увидел как выглядят латино женщины) Там красивые женщины Там не красивые женщины Если бы это был метроном, у него уже стрелка оторвалась
22
35
2
976
13,025
Зачем вообще задавать такие вопросы? Уже столько воды утекло.
Léa Seydoux when asked about who she thinks will be the next Bond girl: "I don’t really care." (variety.com/2026/film/global…)
3
305
Forget-Me-Todd retweeted
LMAO
66
2,338
102
65,515
1,551,122
Это сейчас с каждой игрой так
Fuuuuuuuuuck, you know shit is bad when this dude is making a good point.
1
191
Forget-Me-Todd retweeted
Best part of today is seeing incels across the internet raging that Grand Theft Auto VI is a disappointment because it starts with a mission to pick up groceries and flowers for your girlfriend
4
1,629
516
19,050
499,932