Guido van Rossum: Python and the Future of Programming | Lex Fridman Podcast #341

Lex Fridman and Guido van Rossum dig into the internals of CPython and how Python 3.11 achieved 10-60% speedups through an adaptive specializing interpreter rather than a JIT compiler. They explore Python's design philosophy (readability, indentation over braces), the history and future of static typing via type hints and mypy, and the long-running debate over concurrency, async IO, and the Global Interpreter Lock. The conversation also covers why Python became the dominant language for machine learning and data science, lessons from Guido's time as BDFL, and culture differences across Google, Dropbox, and Microsoft. It closes on a sweeping philosophical riff comparing software layers, DNA, cells, and human civilization as nested self-replicating systems.

• Python 3.11's big speedup came not from a JIT compiler (as many expected) but from an adaptive specializing interpreter.
  01:02:44
• A Facebook engineer built a 'no-GIL' fork of CPython, opening a real path toward free threading.
  02:18:42
• Guido says he is personally not convinced removing the GIL is worth the performance and complexity hit; he likes the GIL as a 'Goldilocks point.'
  02:22:20
• If Python 4.0 ever happens, Guido envisions it breaking the C-extension binary interface (no-GIL) rather than changing Python syntax.
  02:25:47
• Guido admits he held onto the BDFL role too long and was extremely stressed before stepping down.
  02:44:50
• Guido reveals he uses GitHub Copilot every day, calling it a great dance partner despite being 'slightly wrong.'
  03:02:36
• Guido predicts Python will eventually become an invisible legacy layer, like mitochondria in biology.
  03:06:19

• Cited research: an average developer creates 70 bugs per 1,000 lines of code, and 15 of those reach customers.
  00:31:00
• Fixing a bug takes about 30 times longer than writing a line of code, and ~75% of a developer's time is spent debugging.
  00:30:29
• Python's four-space indentation standard was a compromise based on 1980s conventions and typical screen widths.
  00:13:04
• mypy, the original Python static type checker, was created by Finnish developer Jukka Lehtosalo; the name means 'my version of python.'
  01:23:40
• Facebook wrote its type checker (pyre) in OCaml; Google built pytype in Python; Microsoft made pyright.
  01:34:46
• CPython was originally written in about three months, with Guido cutting corners and borrowing ideas because he didn't expect success.
  00:54:55
• At Dropbox, Guido navigated a 5-million-line Python codebase.
  01:46:48
• Microsoft Excel is roughly 35 years old and can still read spreadsheets off old floppy disks.
  02:53:11
• Around the early 2000s it looked like Perl, not Python, might dominate bioscience thanks to its regex engine for DNA search.
  02:35:53
• The Hubble Space Telescope team in Baltimore were early big Python fans in the late 1990s, helping seed scientific Python.
  02:39:36

Books, products and media the guest or host genuinely endorsed here — with the buy link.

Affiliate link — we may earn a commission at no extra cost to you.