PyCon AU 2026
2026-08-26..2026-08-30
Brisbane, Australia
PyCon AU ·
[PyCon AU 2010, PyCon AU 2015 OpenStack Miniconf, PyCon AU 2016, PyCon AU 2016 Science and Data Miniconf, PyCon AU 2017, PyCon AU 2019, PyCon AU 2023, PyCon AU 2024, PyCon AU 2025, PyCon AU 2026]
| title | author | length |
|---|---|---|
| Thursday lightning talks | Daisy Leigh Brenecki, Nic Crouch | 59m |
Forty five‑minute sprint of concrete outcomes: a surprising Python list ordering makes the same loop five times slower; practical comparisons of units… | ||
| Saturday lightning talks | Nic Crouch, Daisy Leigh Brenecki | 01h 02m |
A brisk, eclectic set of micro‑demos and hard lessons: a homebuilt Raspberry Pi + 3.5 RPI LCD cyberdeck that boots, reads books and runs Python after eight… | ||
| Conference welcome (Saturday) | Sophie Quinn-Graham | 12m |
Welcome explains land acknowledgement and reminds attendees of the code of conduct and where to find help on the badge, announces sanitary bins and bathroom… | ||
| Why Your ML Pipeline Needs an Agent (and How to Build One) | Yusuke Shibui, Prashanth Gurram | 31m |
Claims that an agentic ML pipeline reduces months of plumbing to minutes: a demo shows a five‑prompt chain (wine dataset) that produces ~150 lines of working… | ||
| Authentication for humans, why passkeys finally make sense. | George Coldham | 30m |
Claim: replacing reusable passwords with passkeys removes the shared-secret failure mode. Implementation details: devices generate a key pair using public-key… | ||
| The Care of Software: Ethics and Ecosystems | Justin Warren | 39m |
Argues that maintaining software ecosystems by groups matters more than individual artifact creation and maps projects to common pool resources using… | ||
| Adventures in FastAPI: A Choose-Your-Own PyCon AU Talk | Jack Reichelt | 31m |
Live demo shows a FastAPI backend powering a choose‑your‑own slide deck with a React frontend, Postgres + SQLAlchemy for data, and real‑time voting over web… | ||
| Unlocking growth, collaboration and impact through developer communities | Michelle "MishManners" Duke | 30m |
Claims that developer communities drive career growth and product feedback, backed by concrete practices: formal programs like GitHub Stars and Twilio… | ||
| Why Australians don't trust AI and what that means for all of us | Cam Wilson | 44m |
Claims that Australians are high adopters of generative AI—number one per capita for Claude and about half the population using AI tools—but deeply distrust… | ||
| Conference welcome (Friday) | Jack Skinner, Sophie Quinn-Graham, Nic Crouch | 19m |
Welcome announcements list concrete logistics and commitments: venue map, quiet room and gender‑neutral restrooms, catering and bathroom restock changes… | ||
| Ship Reliable RAG: Evals That Grow with Your Retrieval Complexity | Ravi Vats | 30m |
An AI-assisted ticket resolver is built and evaluated across four concrete levels: simple lexical search with BM25 (hit rate ~20%), semantic embeddings (recall… | ||
| Gerrymandering Australia | Rachel Bunder | 29m |
Explains how hypothetical redraws actually change results by building an optimization pipeline using Voronoi tessellation for booth catchments and integer… | ||
| DevRelling in the age of GenAI | Brett Morgan | 30m |
Live demos use Anti-gravity LLM plus reusable skills to generate and iterate Python extractors for large files (a VCV Rack patch and a 15k‑line schematic)… | ||
| Please Don't Put That in a Dict: Better Data Models in Python | Arshia Sharma | 26m |
A hiking-recommendation app built from dictionaries silently suggests closed trails when a field rename turns "status" into None, proving that shape notes fail… | ||
| Productivity: a dialogue | Benno Rice, Daisy Leigh Brenecki | 28m |
Clear examination of whether generative AI actually speeds up work: cites a 2025 result where developers perceive themselves 20% faster but are 20% slower… | ||
| The Good, The Bad, and the Ugly – Python in Secondary Education in Queensland | Matt Collyer | 35m |
Python is common in Queensland classrooms but fewer than 150 schools teach senior Digital Solutions and only about 1–3% enrollment studies it; 97% of those… | ||
| Developing Problem Solvers under Pressure: SREs, DBREs and Ponies | Zoe Braddock | 28m |
Claims that neurorigidity fails under real incidents and that neuroflexibility is the skill to train, backed by eight years of SRE/DBRE on P1 outages affecting… | ||
| Stop Prompting, Start Composing: Agent Skills in Practice | Marty Kieliszek | 29m |
Claims that skills turn capability into competence with concrete gains: modular skill.md folders, progressive disclosure of references, and moving repeatable… | ||
| Automation First, Regrets Later | Alisdair Robertson | 27m |
A platform built and maintained by two engineers manages 25,000+ resources and 65,000+ lines of Python using an Infrastructure as Software approach with… | ||
| Goodbye GIL: Exploring the Free-threaded mode in Python | Adarsh Divakaran | 29m |
Python 3.13 adds an optional free‑threaded interpreter and the presenter builds a benchmark repo and a Flask app, uses virtualenvs, sys.is_gil_enabled and k6… | ||
| Breaking the PR Review Bottleneck | Pavel Oborin | 29m |
Claims: pull request review is the dominant delivery bottleneck and fixing processes before adding AI speeds delivery. Concrete wins include queue reduced from… | ||
| Privacy data encryption & deletion while preserving analytical integrity | XiaoHan Li | 27m |
Claims a practical crypto shredding pipeline that encrypts PII at ingestion with AES GCM, stores per-customer keys in a separate keystore, and implements a… | ||
| From COBOL to Copilots: Why “No-Code” Never Kills the Coder | Sam Bishop | 30m |
Claims that English‑like languages or no‑code will replace programmers get tested against concrete failures: Flow‑Matic and COBOL require experts, HyperCard… | ||
| Testing Beyond Examples: Property Based Testing with Hypothesis | Sam Bishop | 27m |
Introduces concrete use of Hypothesis: property rules instead of examples, defaulting to 100 generated cases and able to run 10,000+ deep CI runs; shows real… | ||
| Feed-Forward, Not Feedback: Rethinking Project Based Assessment | Stephen Tierney | 27m |
Redesign breaks a single end-term project into staged deliverables with feed-forward workshops the week before each deadline, peer critiques scored for quality… | ||
| 30 to 70 PRs a day: How we managed to not wreck our systems | Liz Fong Jones | 42m |
Data and concrete changes decide whether to watch not buzzwords Here is what the session actually claims and shows over eighteen months at Honeycomb a monorepo… | ||
| How to make a blob breathe like a human | Peta Muir | 26m |
An interactive breathing simulator called Blob is implemented in Python Arcade and driven by a parameterized model built from cited human-breathing research… | ||
| Student Showcase | Amanda Hogan, Nicky Ringland, Alison Wong | 01h 34m |
A roster of student projects and concrete outcomes: a working Pygame drone game with score saving, parallax and an accepted visual bug; a Connect‑4 AI that… | ||
| Can I Trust You? Service-to-Service Identity with SPIFFE | Oliver Bassett | 28m |
Shows a concrete replacement for shared long‑lived secrets: SPIFFE identities issued by SPIRE, X.509 certs default to one hour with 20‑minute rotation and JWTs… | ||
| Research Software Stories: Four Years of Lessons in Sustainable Science | Paula Andrea Martinez | 29m |
Four years of collected case studies make a concrete claim: sustained advocacy plus practical incentives moves research software from invisible scripts to… | ||
| The Virtues of Being Lazy | Peter Lovett | 30m |
Describes a concrete claim: Python 3.15 adds a lazy keyword that defers module imports to first use and measurably cuts startup time and memory. Live demo… | ||
| 40,000 CVEs and Counting: What 60 Years of AppSec Tells Us About 2026 | Pedram Hayati | 33m |
History shows concrete shifts and stubborn repeats: 50,000 CVEs now, projecting 290k by year end, active exploitation overtakes phishing, and complexity plus… | ||
| When the demo has to land | Damian Brady | 31m |
Shows exactly what gets built and what actually worked: live demos for Microsoft and GitHub keynotes, a 100‑minute analyst demo distilled from 19 densely… | ||
| Knocking On A Process’s Door: Introspecting Live Python Processes | Raphael Gaschignard | 28m |
Live Python processes get inspected not by vague examples but by concrete tools and demos: cProfile plus flame graphs via snakeviz to show 168s total with 139s… | ||
| Hash It Out: Teaching Password Security Through Code | Alex Penna | 22m |
A hands-on classroom project teaches hashing and rainbow table attacks by having students write Python code using MD5 + hashlib to precompute lists of common… | ||
| Hands-on learning in the age of AI: are developer workshops still relevant | Graham Dumpleton | 28m |
AI answers most technical questions in seconds, so the concrete claim is that hands‑on workshops teach what reading cannot: muscle memory and diagnosing broken… | ||
| From Jupyter Notebook to Eruption: Getting Geohazard Research into Production | Yannik Behr | 28m |
Earth Sciences New Zealand runs a platform that moves Jupyter notebooks from laptops into production via risk-tiered environments and a clear tier progression… | ||
| Kubernetes as LEGO: Building Platforms one block at a time | Dipendra Paudel | 32m |
Shows a concrete path from primitives to a working platform: containerize a Django app, run Postgres with a StatefulSet and PVC, add Celery + Redis and… | ||
| How Being Terrified of Code Helped Me Get Girls to Love It | Gwyneth De Guia | 28m |
Claim that girls leave STEM three times is backed by concrete numbers: only 27% vs 44% of girls versus boys plan STEM electives and roughly a third of girls… | ||
| Assessment in the age of AI | Thea Koutsoukis | 25m |
Survey data show 30 conversations with educators and a 2025 ACARA sample of 4,700 year 10s where 26% use AI frequently, 36% occasionally and only 37% meet ICT… | ||
| Things I don’t worry about as NumPy does them for me | Kai Striega | 27m |
NumPy is a thin header pointing at a flat buffer so a transpose often just flips shape and strides instead of touching elements, while a reshape can trigger a… | ||
| Switching from Sphinx to Markdown | Kattni | 35m |
Argues Markdown is deliverable not just desirable: conversion uses Pandoc plus custom regex scripts, MkDocs (and forks like Proper Docs/Zenscroll)… | ||
| Vibe-Check Your Vibe Code: Surviving the Era of Agentic Open Source Dependencies | Nicky Ringland | 28m |
Claims that agentic code inflates commit volume and reproduces old vulnerabilities with measurable harms: hallucinated dependencies at ~5% commercial and ~22%… | ||
| Experiments in Engineering without reading code | Ben Taylor | 30m |
A project builds a Coding Agent and a software factory around a Python codebase that is never read, using Claude then Opus 5, an Electron wrapper, Axe… | ||
| Why should I write software well? | Tennessee J Leeuwenburg | 30m |
Argues that software quality is a time‑based tradeoff with concrete examples: adopting pydantic and dataclasses versus keeping fewer dependencies, prototyping… | ||
| Why and How to Cite Software (and Get Your Software Cited Too) | Stephanie Chong | 28m |
Shows concrete evidence that software is under‑cited and gives practical fixes: Howison and Bullard 56% uncited and breakdowns of 286 mentions; demonstrates… | ||
| AI-assisted coding, Python and dbt: taking Chronos live in two days | Jay Stein, Sergio Pintaldi | 33m |
A project produces 25 years of gap-free half-hourly weather data in two days by combining Python, dbt and Chronos-2 with Databricks, targeted prompting and… | ||
| Just Enough Reliability | Shaun O'Keefe | 28m |
Argues for prioritising a small set of critical user journeys mapped to concrete endpoints, replacing noisy sitewide alerts (example: 3% errors for 5m) with… | ||
| Whoops! I Accidentally Leaked My Credentials (again) | Eve Martin-Jones | 30m |
New research presents a working pipeline that ingests artifacts from GitHub, npm, PyPI, Maven and Docker, unpacks releases and unpacks and scans each Docker… | ||
| Launching Rockets with Python | Freddy Mcloughlan, Amber Taylor | 31m |
University rocketry team launches several 12ft competition rockets in 2025 and 2026 using a custom Python process manager called the Ground Control Station… | ||
| Your Python container is 90% stuff you didn't ask for | Scott Norris | 30m |
Docker pull python:3.x promises Python and instead delivers ~430 packages — a shell, package managers, compilers and hundreds of known CVEs — so the real… | ||
| Vibing Vision: how we built a pose tracking art installation in two weeks | Charli Posner, Will Cohen | 28m |
A pose‑tracking festival installation gets built in two weeks using on‑device models and runs live at Lost Paradise; the pipeline uses OAK‑D cameras, DepthAI… | ||
| How to give your Python code to someone else | Russell Keith-Magee | 30m |
Concrete options and tradeoffs are described: use PEP 723 inline script metadata or project-level pyproject.toml (PEP 621) so pip and runners know Python… | ||
| Why computer science matters to musicians | Jesse Manser | 26m |
A touchscreen live-rig is built on Ableton with a Python bridge, using Max for Live to access APIs and TouchOSC for custom controls; it runs loopers, a… | ||
| I am bad at things – And that’s okay | Noah Kantrowitz | 28m |
This session claims that concrete practices decide whether teams survive mistakes: checklists for deployments reviews outages that are practiced and blameless… | ||
| From Trauma ICU to InfoSec: A Nurse's Pivot into Cybersecurity | Sam Marsh | 28m |
Claims that nursing protocols map directly to incident response and that the cybersecurity skills gap can be filled by cross‑sector hires, supported with… | ||
| State of Exception(s) | Benno Rice | 26m |
Cloudflare returns many 500s from about 11:20 UTC to 17:06 UTC after a generated data file doubles expected size and exhausts memory causing process crashes… | ||
| Is This Actually Getting Better? A Three-Layer Approach to Measuring AI Products | Wesley Volz | 30m |
Case study of an AI coding agent that actually measures results rather than promising examples: a pragmatic three layers: offline, online, qualitative… | ||
| Enterprise Grade Accessibility for Everyone | Erin Zimmer | 28m |
Practical rundown: enterprise processes, not platitudes. Claims: daily pipeline, VPATs published, 98% of automatable violations caught and a 25% drop in new… | ||
| Is storing files in Slack emojis a bad idea? Maybe. Let’s do it anyway! | Ashley Lamont | 27m |
A file system is implemented by encoding filesystem objects into Slack Emojis and managing them via the Slack API, Python, FUSE, and CBOR2; 116,000 custom… | ||
| "Disentangling formatted strings with T-strings" | Simon Merrick | 27m |
Python 3.14 adds t-strings that evaluate to a Template type not a plain string, exposing an interpolation object (value, expression, conversion, format spec)… | ||
| "from where?!? import what?!?" | Alyssa Coghlan | 29m |
Shows how to find which file actually loads when imports misbehave: run python -m site to print sys.path entries, inspect module and module __spec__ to see… | ||
| "An Intro to Glue Work" | Jade Edward | 31m |
Defines glue work precisely, cites concrete data (women volunteer 48% more, asked 44% more, accept 76% vs 51%), and connects it to AI workflows and tooling… | ||
| "From On-Prem to the Cloud and Back Again" | Luke | 32m |
Claims: FastMail runs about 250 servers across three datacenters storing ~5 PB for ~200k paying customers, operates ~100 services with 5–10 prod deployments a… | ||
| Not Black and White: Cryptic Crosswords, Code Golf, and the Rules of a Language | Amanda Hogan | 30m |
Shows how cryptic crossword techniques map to real Python tricks and then shrinks a solution from a classroom 211 characters down to 55 characters using… | ||
| The 20-Year Port: from PhD project to polished professional Python packaging | Paul Leopardi | 29m |
A twenty‑year MATLAB geometrics toolbox becomes a production Python package by porting code, tests and docs, using coding agents (Google anti‑gravity harness… | ||
| Your Research Code Works.. Until It Doesn’t: Rethinking Reproducibility | Zara Hassan | 31m |
Claims that reproducibility failures accumulate as Reproducibility Debt rather than erupting suddenly and that small shortcuts (missing deps, undocumented… | ||
| From Pixels to Points: Inferring Archery Scores with Computer Vision | Jacob Seiler | 28m |
A working pipeline infers arrow tips and scores from photos with 3,500 annotated arrows and about 504 ends, using YOLO oriented detection plus a second pose… | ||
| "TOGA: Checklists and Safety Measures for "Yoloing" in Prod" | Chelsea Finnie | 28m |
Covers concrete rules and failures: a real incident generates 50+ checklists, debris severs 650 wires and forces a rethink to “what’s still flying,” landing… | ||
| "Optimizing Scientific Python with C++, CUDA, and Serverless Compute" | Alex Zinovev | 31m |
Migration story with numbers and tools: moving algorithms that touch tens of millions of points into serverless where Lambda has ~10 GB RAM and short timeouts… | ||
| "The Neurospicy Toolbox: A Guide for Educators" | Maddie Mackey | 29m |
Practical, concrete toolbox: step‑by‑step strategies for designing coding projects for neurodivergent students, backed by Australian figures (about 290,000… | ||
| "What Happened in Production?! Instrumenting with OpenTelemetry" | David Bell | 28m |
Daily 2pm crashes turn out not to be a mystery but an OpenTelemetry Tracing story: Python auto‑instrumentation reveals DB deadlocks that metrics and logs miss… | ||
| "Conference welcome (Thursday)" | Jack Skinner | 12m |
Contains concrete logistics and community rules: A core network for wifi with on‑screen access prompts, a badge map and three‑word game, location for swag and… | ||
| "Snek Game Finals" | 06m | |
A live Python snake game runs on an ultrawide screen between sessions and the three players with the top three scores Alyssa Jay and Bella go head to head in a… | ||
| "Closing address" | Jack Skinner, Nic Crouch, Sophie Quinn-Graham | 23m |
Logistics, prizes and community next steps get spelled out: development sprints all day in the Leon room with catering and coffee, extra stickers set out on… | ||
| nothing matches | ||