Not Just Python

Keynote delivered at PyCon Colombia, July 24, 2026, in Medellín

Transcript

1. Not Just Python Luciano Ramalho luciano@ramalho.org https://codeberg.org/ramalho/pycon-colombia-2026
4. Lightning bolts at Port-la-Nouvelle, southern France • Photo by Maxime Raynal Creative Commons Attribution 2.0 • From the Wikimedia Commons
5. Sets in Python
12. Python solution #1 We write Q ⊆ D as Q <= D Also: Q.issubset(D)
14. Python solution #2 We write F ∖ C as F - C Also: F.difference(C, ...)
18. Visible set demo • Repository: https://github.com/ramalho/visible-set-demo
21. Sets in Python • Covered in Fluent Python, chapter 3, “Dicts and sets” • Algorithms and hash tables detailed in https://www.fluentpython.com/extra/internals-of-sets-and-dicts/ – How efficient are Python’s dict and set? – Why can’t we use any Python object as a dict key or set element? – Why are set elements unordered? – Why and how do dicts preserve insertion order? (since Python 3.6)
24. Photo by Maxime Raynal • Creative Commons Attribution 2.0
25. The Lizard Book
27. Principles of Fluent Python (1) • Target audience: professional devs using Python • Focus on special methods (__dunder__) from chapter 1 • Functions in depth before classes • More than 1000 external references – with short URLs that I can update when necessary • Opinionated essays about language design (Soapboxes) • Theory illustrated with practical examples
30. Principles of Fluent Python (2) • Leveraging what exists before (re)inventing – • Native data structures in depth Contrasting concepts (covered separately in the docs) – Lists, tuples, arrays, queues – Data classes, named tuples – Threads, processes, coroutines ☠☠☠ • Balanced coverage of all supported typing disciplines • Metaprogramming and metaclasses!
32. The lizard’s journey (so far) • 2013: contract signed with O'Reilly • 2015: Fluent Python released (EN, PT-BR) • Through 2022: 1st edition published in 9 languages • 2022: Fluent Python 2nd edition released • 2024: full text at PythonFluente.com (PT-BR); Brazilian publisher drops 2 nd edition • 2025: crowdfunding for the 2nd edition in PT-BR • 2026: delivering to crowdfunding backers; available in bookstores
35. Woodcut by Thiago Castor Traditional technique and tools used to make cover art for popular literature (literatura de Cordel)
37. Python Fluente, 2ª Edição (2023) • Coded with Asciidoc (supported by O’Reilly) • PDF and HTML generated with Asciidoctor.org (Ruby) • Translated from English to Portuguese by a human: my friend Paulo C. Oliveira • Revised with the help of the community and languagetool.org – not an LLM: a narrow AI specialized in spelling and grammar correction
38. Photo by Maxime Raynal • Creative Commons Attribution 2.0
39. The Typing Map
42. Photo by Maxime Raynal • Creative Commons Attribution 2.0
43. Making Sense of AI
46. Evil tech bros • Altman, Musk, Pichai, Zuckerberg, Amodei etc. promise AGI for the benefit of mankind • But they stole everybody’s content • Abused human data workers • • Unleashed unsafe “companions”, leading to multiple cases of self-harm Want to cover the Earth with data centers (Sutskever)
47. Large scale misinformation • Confusing narrow x broad AI applications • Hyping existential risk to distract from present harms • Circular financing: – • I give you $100B, you buy $100B of my stuff, now we both have $100B, total $200B, free money! Accusing the Chinese of stealing IP while profiting from the largest IP theft in history
48. AI hype in hyperdrive • Autonomous cars are not really autonomous • Humanoid robots are useless, fake, or frauds – • • Read Rodney Brooks, professor of robotics at MIT https://dub.sh/gUwas1F But Tesla now is all about autonomous vehicles and humanoid robots! And SpaceX is now all about AI and datacenters!
49. Our context is very particular • • LLMs are useful for software development We have tools and methods to verify generated artifacts, but most professions don’t – Automated testing is essential • • But LLMs may cheat, writing tests that pass even when the implementation differs from the spec Implement TDD: tests written first, agent can only update tests with permission
50. My status • Member of the class action “Bartz v. Anthropic” – I may get $ 1.500 in damages for Fluent Python • Subscriber of Claude Code at $ 20/month • Having fun using Claude Code for hobby projects • • I would be very stressed if I was in a company and was forced to do agentic coding I predict a major financial scandal (like WorldCom)
51. Moral dilemmas • Cognitive dissonance – • • Remembering Aaron Schwartz, destroyed by the Obama DOJ for downloading articles from JSTOR I stayed away from AI until I decided to keep my connection to my friends in the professional software development community who have no choice Up next: explore local models and open source models (harness > model)
52. The Knowledge Smoothie • • • • • SAM ALTMAN: Let’s make a smoothie! Who can contribute with fruits? EVERYONE ELSE: Great idea, here’s the fruit I have! SAM ALTMAN: The smoothie is ready. It’s $20 a glass! EVERYONE ELSE: But you took our fruits to make it! SAM ALTMAN: No I didn’t! Where’s your fruit? Show me!
53. The solution • Any LLM vendor who refuses to identify every external source used in training must open source the model. • Too radical, or just fair and common sense? • Civil disobedience: until that happens, let’s ignore copyright!
55. Photo by Maxime Raynal • Creative Commons Attribution 2.0
56. The Power of Communities
57. © All rights reserved by Kushal Das
58. How the Python community is awesome • Built around a language designed for “end user programmers” – • • • Scientists, analysts, journalists, artists etc. Transparent and representative governance Not dependent on any specific government or corporate support Not driven by profits
59. • • • Guido van Rossum: a founder who truly cares about inclusivity “Python is for girls” t-shirt at PyCon US 2012 Q&A: for women only (for balance)
61. How Pyladies are awesome • Driven by solidarity • Grass-roots, bottom-up organization • Location based • Not driven by profits
62. Local Communities #IRL
63. We need real connections, face to face • • • Not driven by consumption, companies or hierarchical institutions Online “social networks” are no substitute for that – Too much algorithmic meddling – Too many trolls and robots Face to face collaboration may be the best antidote to culture wars and radical polarization
64. Documentary: Torrow (Demain) • Directed by Cyril Dion & Mélanie Laurent • Grass-roots, bottom-up projects • Location based – Local currencies – Community urban farms – Community seed banks
65. Location based, not driven by profit • Hackerspaces (by definition) • Martial arts dojos, capoeira schools (sometimes) • Samba schools (dancing, music, sewing, building) • Bicycle repair workshops • Community vegetable gardens • Hobby clubs: model planes, model trains etc.
79. Photo by Maxime Raynal • Creative Commons Attribution 2.0
80. Idea: localized online messaging • • • • To create a community, 3 founders must be together in the same place To join, new users must meet 3 existing users in person Each month every user must meet 3 other users in person, otherwise access is blocked To unblock, just go meet 3 other users in person “3” is just an example to get started
81. Benefits of localized online messaging • Supports the habit of leaving the house to see people • No bots • Trolling has real life consequences • Enables local solidarity: – Who has a power drill that I can borrow? – Who can help me fix a bicycle tire?
82. Photo by Maxime Raynal • Creative Commons Attribution 2.0
93. Enjoyed the talk? Scan the QR code and share your feedback. Your input helps us improve future sessions! QR CODE IMG