Patrick Collinson on programming languages, AI, and Stripe's biggest engineering decisions (AI column)

I saw something interesting this weekend. interview Patrick Collinson, co-founder of Stripe and an enthusiast of programming languages. He was interviewed by Michael Truwell, co-founder and CEO of Anysphere, the company that created Cursor. The guys discussed a wide range of topics, including the future of programming languages, development in general, how things are at Stripe and beyond:) The main ideas, as usual, are below.

1. Experiments with Smalltalk and Lisp Patrick's first startup was written on SmalltalkIt offered an interactive development environment with the ability to edit code in real time. Hiring employees wasn't a problem - smart people learn new languages quickly. On Lisp, Patrick made a bot for MSN Messenger using a Bayesian predictor. (He could talk to people.). He also experimented with genetic algorithms to optimize keyboard layout. (Confirmed the effectiveness of Dvorak). But all these side projects were far from the current capabilities of AI. 2. Philosophy of programming and development environments Patrick advocates combining runtime, text editing and code execution environment. As an example, he mentions Mathematica. (It's just now Wolfram). Patrick wants the IDE to see profiling information when hovering over the code, as well as logs and error messages. Here Patrick remembers another non-profit company DynamiclandIt's about spatial computing. He admires the work. Bret Victor. This is a company, but he's not sure that this approach applies to arbitrary systems. 3. Technical solutions in Stripe Here's Patrick discussing the impact conway It focuses on architecture, emphasizing the importance of APIs and data models in shaping an organization. Strangely, Patrick recalls the iOS ecosystem and says it outperforms Android due to better API design – to me, the iOS ecosystem looks good to the consumer, and engineering is just the bottom. 4. Choice of technology MongoDB and Ruby remain fundamental Stripe technologies 2010 year (||This is the legas of technology for me.||). However, the guys rewrote some services in Java to improve performance. I wonder what in 2022 A large-scale project on the unification of abstractions Stripe API v2, which is now gradually rolled out on the market, has started. Patrick explains the complexity of large migrations. 5. The Future of Programming with AI Patrick uses LLM mainly for factual questions and programming through Cursor. And then Michael shared his vision, which develops Cursor Programming languages can become less formal and more result-oriented AI can help refactor codebases and improve architecture Cursor plans to create a more integrated development environment with AI in the background 6. Economic aspects and research on progress Here the guys discussed different theses and opinions, it’s funny that they touched the whitepaper.Measuring the Impact of Early-2025 AI on Experienced Open-Source Developer Productivity"But Patrick hasn't had time to study it yet." (I have already studied and disassembled for you: 1 and 2). Summarizing their thoughts, there is no clear evidence of performance gains from LLM. 7. Progress Research and the Example of Biomedicine Patrick believes that the need for research on progress has increased due to increased degrees of freedom. For example, Patrick co-founded. Arc Institute into 2021 A year for biomedical research. They are working on basic models for biology using DNA, and are also trying to create a virtual cell to understand complex diseases to fight complex diseases like cancer and neurodegeneration.

All in all, this interview seemed very interesting to me - many futuristic topics and ideas were touched upon, and the mundane issues around Stripe and Cursor were addressed.

#AI #ML #Management #Leadership #Software #SoftwareDevelopment #Architecture #Processes