-
Notifications
You must be signed in to change notification settings - Fork 14
20 Aug 2025
Philipp Ahmann edited this page Aug 20, 2025
·
4 revisions
** indicates TSC voting members*
- *Philipp Ahmann
- *Nicole Pappler
- *Luigi Pellecchia
- *Alessandro Carminati
- *Matt Weber
- Pete Brink
- Simone Weiss
- *Gab Paoloni
- *Paul Albertella
- Pierre Brangier
- Henrik Brändle
- *Chris Temple
- *Olivier Charrier
- Mohammad Abdoli
- Kate Stewart
- Sudip
- *Luigi Pellecchia
- *Gab Paoloni
- *Igor Stoppa
- *Paul Albertella
- *Christopher Temple
- *Nicole Pappler
- *Olivier Charrier
- *Kate Stewart
- *Paul Albertella
- *Igor Stoppa
- *Alessandro Carminati
- *Gabriele Paoloni
- *Matthew Weber
- *Matt Kelly
- *Sudip Mukherjee
- *Lukas Bulwahn
- Henrik Brandle
- Pierre Brangier
- Patrick Uven
- Sebastian Hetze
- Panos Kalorog
- Pete Brink
- Michael Mahoney
- Jaylin Yu
- Victor Lu
- Naresh Ravuri
- Elana Copperman
- Steve VanderLeest
- Qasim
- Stephen Oresanya
- Action items in github issues
- no currently pending AI
- Booth representative -> Please fill in, when you have some time in your schedule. Link shared via mail and in chat.
- Few slots still open. Thanks to all who already signed up for supporting!
- Hold for a community happy hour on Tuesday after Tux Trek
- Use discord channel for alignment
- Discord channel exists for OSS EU
- Date considered: Nov 18th to 20th
- Olivier and Gab are checking (due to vacation period). Philipp need to check room availability, but it works in principle at ETAS Hildesheim
- More details on topics to be discussed at coming TSC meeting.
- Should fix location latest by September 17th
- See: https://github.com/elisa-tech/directory/pull/10
- Please continue to review!
- Usage of AI in the Linux Kernel development process (from requirements to testing) to reduce the risk of systematic error
- Existing ksummit thread on kernel mailing list:
- Annotating patches containing AI-assisted code: https://lore.kernel.org/ksummit/[email protected]/T/#t
- The role of AI and LLMs in the kernel process https://lore.kernel.org/ksummit/49f1a974-e1e6-4be5-864e-5e0f905e1a8f@paulmck-laptop/T/#t
- No functional safety standard that regulates the AI for the creation of software development process artifact, like generating requirements or test cases or architectural diagrams.
- Today you can ask AI to create specifications based on existing code. But the question is: How can you reduce systematic errors by the use of AI?
- Birdeye view idea for a virtual cycle: Let create a specification. Than take the specification to create test cases. Then apply the test cases to the code. Take iterations until code, spec and tests are consistent.
- Be careful not to miss the verification measures on the left side of the v.
- Not replace existing verification model, but plug in AI to reduce systematic errors and have something equivalent to the requirement ISO 26262 put on the development cycle.
- You can train AI with requirements and providing the code and this creates very good diagrams from first experiments. (Doesn't replace human though but accelerates discovery/clarification when learning a complex area of code.)
- Continue this discussion during next TSC.
- Recently auto transcription was switched on in a working group meeting.
- This was an automatic transcript. So person had to exit and rejoin to turn it off.
- Also in Workshops this happened in the past.
- Any concerns or viewpoints on this topic?
- Note: In the past we explicitly decided not to record meetings, but we never talked about transcripts nor AI summaries.
- The Organizational Rules currently state: "No recordings of working group meetings are permitted. Special provisions may be arranged for recording in advance with explicit consent of the participants."
- Concern on capture the conversation with the names as transcript.
- Chatham House rule typically followed during meetings, so transcripts may violate with it.
- It can be useful to have recordings for presentations.
- AI is not trained for our safety language. So transcripts easily are wrong.
- Transcripts need to revisited.
- For workshops transcripts could be beneficial for better following the conversation. It is also an inclusive element for non-native speakers. It can be used as an aid for some.
- Rather than principle denial or a simple "against/for" we should have some guiding rules.
- Make sure that all participants agree in e.g. record or enable transcripts.
- In general recording and transcriptions will not be turned on.
- Properly document the agreement in the meeting minutes
- In whatever is agreed transcripts must not be shared and are only for the personal use.
-
AI-Philipp: Draft a revised policy and put it under TSC GitHub
- Philipp will prepare a draft and create a PR with Paul, Matt and Olivier as reviewers.
- Kate is currently checking with LF Research on possibilities and if there is overall benefit seen in such a survey.
- Continue the discussion also in person during OSS EU in Amsterdam (potentially with the LF Research people like Hillary) and also during TSC meeting.
Up to 3 bullet points
-
Aerospace/Space
- It was a slow summer with vacations. AeroWG is working towards restarting whitepaper and use case topics in Sept.
- Use case testing group (wkly) - Maturing documentation for the two demos. Investigating a cFS use case demo (have not captured "what" this would be yet.)
- Automotive
- Medical (paused)
- Architecture
- Linux Features
- OSEP
- Tools
- Systems
- GitHub Enterprise for ELISA -> Upgrade proceeded 😉
- Mohammad Abdoli joined first time. Coming from SEGULA consulting Volvo Trucks.
- To be defined: SDV seminar by EMQ to motivate the update and cloud part of HPC systems
- RT webinar once PREEMPT_RT as last patch is mainline.
- Past webinars on website: https://elisa.tech/seminar-series/
- Later this year:
- ELISA user story by a company.
- Julia Lawall about formal verification (derived from Lund Linux Con presentation)
- 25-27 Aug - OSS EU
- 09-14 Sep - IAA Munich
Collaborative editing: https://mensuel.framapad.org/p/elisa-tsc-minutes-a9tp?lang=en