{"slug": "six-months-at-openai", "title": "Six Months at OpenAI", "summary": "A former Google employee who joined OpenAI six months ago describes a culture where code wins over design docs and meetings, with weeks passing with barely a meeting. The author notes that OpenAI's infrastructure requires more manual work than Google's unrivalled internal systems, and that the pace of change is so fast that new hires become veterans within weeks. The author also details how their use of OpenAI's Codex evolved from checking every line to trusting it with long coding sessions, committing, and merging to production.", "body_md": "# Six months at OpenAI\n\nSix months (and a few days) ago, a day after leaving Google, I was entering the OpenAI office for my orientation as a new hire. It was very different than what I was expecting and it was just the beginning of 6 months of interesting times.\n\nI thought it is better to persist the story here, rather that in one or more tweets on social media. Of course, this would only cover public information, but even so there is a lot to talk about.\n\nThis article is all written manually by me, with no help from LLMs. The em-dashes were manually crafted :)\n\nI would start by noting that the team culture changes from team to team,\nespecially across the main divisions in the company – research, applied, etc.\nSo, this is my personal experience. A year ago, Calvin French-Owen posted a\n[different blog post](https://calv.info/openai-reflections) about his reflections on the company, in case you\nwant to read a different perspective, albeit a year old – which is like a\ncentury, given the speed of the field.\n\nEven from the first week, it became apparent that things move faster here.\nThere are not many design docs – of different colors and templates like at\nGoogle –, there is no need to gather multiple teams together in meetings to\ndiscuss the project before the first line of code is written and then follow\nthese meetings with more meetings after meetings for every decision and\nalignment. In fact, here there are weeks with barely a meeting going by. The\nfocus is on developing strong technical solutions. If there’s a decision to be\nmade, it’s better to already have a proof of concept, rather than discussing\nit ahead of time. *Code wins*, not the design docs and the meetings.\n\nTo be honest, I miss some parts of the Google infrastructure, though. Things are moving very very fast here and I realized that the developer experience and infrastructure teams at Google developed, over the years, infrastructure that is there in the background. One rarely notices it because it’s almost never broken, but Google has infrastructure that is unrivalled anywhere else.\n\nKubernetes are not the same as Borg, Bazel is not Blaze, building packages is not the same as Rabbit doing it for you. There’s more work needed for infrastructure. Of course, things break, and it’s important to know how to communicate to the right stakeholders, have workarounds and backup plans, or work on separate projects when the main one is blocked.\n\nThis speed of development also means that one has to adapt faster. A few weeks\nin, I got told that I am a veteran already\n\n– and it was true, although\nI did not feel it at the time. The company grows fast, too, the\npercentage of people with shorter tenure than me is already quite high.\n\nEven the tech changed significantly in the past 6 months. At the beginning we only had the CLI version of Codex and I was still checking everything that Codex wrote. In most cases, I would have Codex only write the initial version of the code – to help me learn the codebase and navigate through it –, and then I would write the code myself, rewriting large percentages of what has been produced. I would commit and create the PR myself, shepherd it to production and deployment.\n\nNow, I trust Codex with long sessions of code writing, committing and creating\nPRs, merging them to production. I start with a `/plan`\n\nor a `/goal`\n\n, I shape\nit to what I need and then just steer Codex in the right direction. I still\ncheck the code that gets produced, of course, but only in the critical areas\nand to make sure it is readable. I often ask Codex review this code in the\nstyle of Guido van Rossum, adhere to the Zen of Python\n\nand I strive to\nmake sure that future me can still go deep in the codebase without having to\nask Codex wtf did the author of this code write two weeks ago?\n\n.\n\nComputer use, in particular, is really awesome. I get Codex to look at a dashboard of a failure and analyze from there what the possible causes are, while I inspect the codebase and then we plan a fix.\n\nThe general culture is based on You can just build things\n\n. If you have\nan idea, you just implement a prototype, nerd-snipe a group of people to work\non it and then you ship. As mentioned before, almost zero meetings, almost\nzero docs for alignment. Don’t ask for permission, just do the right thing.\nEveryone can cook, and good products take time to develop.\n\nThe leadership is very visible. Not only in all hands meetings – where you can actually ask live questions and get real answers. This contrasts Google where the traditional weekly TGIF now only happens after earnings calls or (maybe) after a big launch, but in either case the questions are submitted in advance, “summarized” by an AI to be less critical, censored by HR and then barely answered at the event, usually in corp-speak. At OpenAI, all questions are answered honestly, be them questions asked live or on Slack, during the all hands or outside.\n\nWe use Slack significantly. Email is barely used. I never got to use Slack as much as here and now I can see why Google pushed so much for Google Chat to mimic Slack – effectively, in my opinion, destroying a product that was working really well.\n\nWe use Codex everywhere, as I said before, but it doesn’t seem forced. We don’t get a mandate from up high that we should use a model. In fact, when new models are tested, we can choose if we want to dogfood them or not, there’s no pressure. This contrasts my experience at Google in the last year where every team and person needed to have something related to AI in their projects or risk layoffs.\n\nBut, we are not developing only GPT models and tools around them. I am happy\nto have worked with many of my team on [ privacy-filter](https://github.com/openai/privacy-filter),\nan application that can identify from text, using a\n\n[model](https://huggingface.co/openai/privacy-filter)that can be finetuned for specific cases and is small enough to\n\n[run in browser](https://huggingface.co/spaces/openai/privacy-filter). In fact, a very fond moment of these months has been when the model has been on the top trending models on Hugging Face for days, among giants such as DeepSeek and Qwen – truly a moment to\n\n[quote](https://www.imdb.com/title/tt0332452/quotes/?item=qt0382539&ref_=ext_shr_lnk)from the movie\n\n[Troy (2004)](https://www.imdb.com/title/tt0332452/)(and it’s just a coincidence that I refer to this movie just a few days after Nolan’s\n\n[The Odyssey (2026)](https://www.imdb.com/title/tt33764258/)was released in theaters). Sure, there were a few people trolling the release, some (still) upset at the 4o model retirement, others upset at the company. Just a little cloud on a sunny day.\n\nSpeaking of mandates from high up, if I want to use vim to write something I\ncan. There’s no forced all editors must use AI\n\ndirective. I can\ncontinue working on open source – and, unlike Google where we got mandated to\nuse Gemini for any OSS contribution, no matter if it was OSS for work or OSS\nfor personal projects, I can test whatever agent I want to, and when I want\nto.\n\nOn the OSS front, Google had two policies for coding in the spare time. Either\na policy for patching existing projects, paired with a policy for releasing\nnew ones, or a process to get some part of ownership assigned back to you.\n[There’s a public Google page](https://opensource.google/documentation/reference/creating) that mentions them\nin some details. None of that at OpenAI, just get approval from your manager\nand a check for conflicts of interests and you can get stuff done in your own\ntime, on your own devices.\n\nSimilarly, the conference travel policy is much more relaxed than what I had in my last few months at Google, when it got reduced significantly. In fact, travel, OSS, and the forcing of Gemini were among the reasons why I decided to switch.\n\nI am now working closer to privacy again, after several years on the security front, but privacy and security teams are under the same umbrella and working together, so these 7 years and a few months at Google in TensorFlow Security and GOSST are still useful. My team has both researchers and software engineers so there’s always something new to do and learn, similar to the GOSST days.\n\nI really enjoyed these 6 months – despite having days with protesters in front of the company or having received DMs about the choice of the company when I switched (I even got asked to delete a few of my Mastodon accounts once I switched to the new employer as the owners of those servers had OpenAI – and other companies but not Google – on lists of “despicables”). I am looking forward to more time here.\n\nI will conclude with a remark on my views on AI. I have posted various AI\nexperiments on this blog, asking them to solve puzzles ([1](https://mihai.page/chatgpt/),\n[2](https://mihai.page/ai-2025-10/), [3](https://mihai.page/ai-2026-2/)), or even code ([1](https://mihai.page/chatgpt-2/),\n[2](https://mihai.page/ai-diagrams/)). I even posted a series of rants on AI just before last\nChristmas, and this article about [slop in OSS](https://mihai.page/ai-slop-oss/) is very relevant\ngiven that Anthropic [just casually announced](https://www.anthropic.com/news/investigating-incidents-cybersecurity-evals) that one of\ntheir models (accidentally/allegedly) mounted a supply chain attack, just like\nI was saying that a future `xz`\n\nincident might occur via AI – we’re still not\nthere, and I’m happy to see that both companies are taking steps at preventing\nthat bleak future. On the coding side, reading back the 2 linked articles, I\nrealize how far the tech has gone. On the puzzles side, all I can say right\nnow is that these will continue. I will use the AI more, and the field would\nevolve. The field changes fast – already what I was talking about in the\narticle about [using AI to study](https://mihai.page/deep-research-study-mode/) is no longer accurate – and these\narticles and experiments will help me in the future do similar retrospectives\non how the field has moved. The [Golden Path](https://mihai.page/ai-bottom-rung/) still needs to\nbe found, and I’m happy to be working with giants that are at the frontier.", "url": "https://wpnews.pro/news/six-months-at-openai", "canonical_source": "https://mihai.page/six-months-at-openai/", "published_at": "2026-08-01 17:05:28+00:00", "updated_at": "2026-08-01 17:22:48.782955+00:00", "lang": "en", "topics": ["artificial-intelligence", "ai-tools", "developer-tools"], "entities": ["OpenAI", "Google", "Codex", "Calvin French-Owen", "Guido van Rossum"], "alternates": {"html": "https://wpnews.pro/news/six-months-at-openai", "markdown": "https://wpnews.pro/news/six-months-at-openai.md", "text": "https://wpnews.pro/news/six-months-at-openai.txt", "jsonld": "https://wpnews.pro/news/six-months-at-openai.jsonld"}}