OpenTeams

10/14/2026

Throwback to PyTorch Conference 2025.
Guilherme Leobas shared how CPython’s own tests help make Dynamo correct. This year he is back with the next chapter. Watch the throwback and visit his poster on Oct 20 at the Community Expo.

10/12/2026

Good AI work starts with shared context.

Travis Oliphant, founder and CEO of OpenTeams, shares his view in our latest whitepaper. Open standards grow when maintainers, researchers, and teams all help shape them.

Read the whitepaper, then tell us what your team would want from a shared standard.

https://github.com/openteams-ai/inthub-whitepaper/blob/main/whitepaper.md

10/10/2026

What works for one developer doesn't always scale for a team.

OpenTeams CTO Dharhas Pothina explains why teams need an easier way to store, share, and distribute environment specifications across their workflows.

Watch Dharhas discuss the challenge at The Python Exchange. https://www.youtube.com/watch?v=rCa3zztvbJk

10/08/2026

You don't choose every risk in your supply chain. Some of it is 15 years old and still there. Brandon Geraci explains. Watch the full talk: on our YouTube Channel (link in bio)

10/08/2026

In science, the ability to reproduce a result is essential. But reproducibility alone does not establish that a result is true, or that the conclusions we draw from it are justified.
Every computational model rests on assumptions. Every scientific claim depends on evidence. And somewhere between the two lies a profoundly human responsibility: deciding whether what we know is sufficient to support what we claim.
I'm looking forward to a talk exploring how Python, knowledge graphs, and OpenMBEE's Flexo MMS can make these connections explicit, traceable, and open to scrutiny.
We'll begin with a satellite attitude-control system, but the questions extend across scientific computing, machine learning, and AI.
The remarkable power of computation is that it allows us to explore what follows from our assumptions. The responsibility of science is to ask whether those assumptions describe the world.
Save your spot: https://tinyurl.com/mrtx5hbs

10/07/2026

Sometimes the best engineering decision is deciding not to build.
During her OpenTeams internship, Vaidehi Pandya learned when determinism beats AI, when automation creates more problems than it solves, and when restraint is a technical skill.
https://openteams.com/guidance-is-not-a-boundary-four-lessons-from-a-security-internship/

Eleven weeks, four recurring lessons: why guidance isn't a boundary, why reasoning isn't verification, and why the hardest call is often choosing not to build.

10/06/2026

Have you ever tried using a model to sort through millions of small decisions?

Brandon Geraci faces this problem while building Artifact Keeper, an open-source artifact manager that checks packages before they reach a team's registry.

Rules can handle the clear cases, but the more complex ones need to enter a review queue that is too large to process manually.

Brandon needs a model that can sort this queue using predefined labels and confidence scores.

Jev is designed for this kind of task. It chooses from predefined answers, such as yes or no, and reports how confident it is.

In the article, he tests Jev, Laya, CLM-8B, and Claude Haiku 4.5 on five package-curation tasks to see how each approach performs.
https://tinyurl.com/25nmunsu

10/05/2026

As more AI tools become available for Jupyter, keeping track of them has become difficult.

Michał Krassowski introduces the Jupyter AI Awesome List, a curated map for users and contributors.

The article explains how AI tools are being integrated into Jupyter and what to consider when choosing one.
https://tinyurl.com/25tp4rp2

10/01/2026

The Jupyter community is gathering in Prague. 🇨🇿

Mike Krassowski will speak at the Jupyter Mini Summit at Open Source Summit Europe on October 6.

His talk covers what's new in JupyterLab 4.6. It was one of the largest minor releases yet. The new features also reach Jupyter Notebook 7.6 and JupyterLite 0.8.

He will also walk through the tools that helped contributors move faster. Other open source projects can reuse these tools and ideas.

Come meet the people behind the work, ask questions, and share your own ideas.

Pre-registration is required. Add the Mini Summit to your OSS Europe registration: https://osselceu2026.sched.com/event/2RjI5

View more about this event at OSS + ELC EU 2026

09/30/2026

You can't drive AI like software. You have to ride it like a horse.

Our engineer Amelia Thurdekoos built a full customer site in a week using Claude Code, then wrote up exactly how she kept every public claim honest along the way.

Read the full breakdown here: https://openteams.com/you-dont-drive-ai-you-ride-it/

AI is stochastic, not deterministic. You don't drive it, you ride it: make documents the memory, use the variation to prototype, and verify every public claim.