Your Kitchen, The New Conference Room
Last week, Stripe announced that their next engineering hub will consist of an entirely remote workforce. Stripe’s existing engineering hubs are San Francisco, Singapore, Dublin, and Seattle. The blog outlines their reasoning, stating:
We are doing this to situate product development closer to our customers, improve our ability to tap the 99.74% of talented engineers living outside the metro areas of our first four hubs, and further our mission of increasing the GDP of the internet.
Teams with remote members often need to add additional rules to their communication to make sure everyone is in the loop. A surprising amount of decisions are made at the water cooler, and this shift will replace that with a Zoom link.
Look — at face value, I get it. Living in San Francisco is not a prerequisite for knowing how to make great software, and the cost of living is prohibitively expensive for anyone who values a saving rate higher than 0%. On the other hand, companies announcing a shift to a remote workforce and then rolling back the decision is not uncommon. Company culture is hard. It’s even harder when your workforce is geographically separated by thousands of miles and several time zones. Relationships aren’t made over Slack and email, they’re made during the walk to lunch or sitting in a room and making small talk. If you can choose between the comfort of your home, or turning on a Webcam to drop in on a coworker, I would guess that the former is more probable. That’s not to say it will never happen, just that it will happen less than a traditional office environment.
The opposite extreme of this seems to be a place where everybody lives and works. Most startups begin this way. Cofounders either live in the office, or turn a shared apartment into a workspace that also happens to have mattresses. As the organization expands the office grows with it, but much of the initial quirks remain.
Stripe is betting that the wave of new productivity and communication tools are an adequate substitute for face-to-face interactions. For engineering tasks, they are probably right. But scaling meaningful human interaction is much harder.
A funny and insightful story from an early Youtube engineer on their plot to kill Internet Explorer 6. Shortly after the Google acquisition, the team used a backdoor called ‘OldTuber’ to bypass the standard production release guidelines. Anecdotes like this help to illustrate the power that developers often have in shaping large scale decision making, especially in popular consumer apps. Link.
A great list from Patrick Collison on projects that were completed rapidly. Apollo 8 took 134 days to launch to the moon. Link.
Google’s new feature that will allow the deletion of your location data. Still no way to prove that the data is deleted, plus the trained models will remain in production. Link.
Blog post on SpaceX’s recent fundraising efforts, and the competition from Blue Origin. Link.
Story of Amazon Prime’s inception with plenty of quotes from the team. It wasn’t an obvious decision at the time. Link.
OpenAI releases a larger version of the GPT-2 language model. This version is still less than half the size of the full model (~345M vs 1.5GB), but shows significant improvement over the original release. Link to paper with updated release guidelines. Link to Colab notebook to try out the text generation for yourself.
🐦 Tweetstorm of the Week
Lex Fridman interviews Oriol Vinyals, a DeepMind and former Google Brain researcher. Discussion on Starcraft, language models, the Turing Test, and AGI. Link.
Not a subscriber yet? Sign up here.