If "DOE has already shifted a large share of funds away from epistemically-sound basic research over to risky bets on LLMs." that funding is not available to PETSc developers. This makes it difficult for PETSc developers to compete for future DOE funding related to LLMs, as we cannot demonstrate much prior work due to the lack of funding from DOE to work with LLMs. The most effective tools for "writing documentation, new tutorials, and refining material for hands-on in-person or virtual tutorials" revolve around LLMs. Ignoring those tools, simply because they have lots of bullshit hypre associated with them, will slow down our ability to improve the documentation, tutorials, and related materials.
On Sep 21, 2025, at 8:36 PM, Jed Brown <[email protected]> wrote:
I think we should apply for a topic that is *NOT* LLMs [1]. It could sponsor a stipend for a grad student to write documentation, to write new tutorials (maybe poll for which topics are in high demand), or to refine material for hands-on in-person or virtual tutorials.
[1]: DOE has already shifted a large share of funds away from epistemically-sound basic research over to risky bets on LLMs. If PETSc were to apply for that one thing, it would send a message that we don't think DOE has gone far enough. I think that's 100% backwards.
Barry Smith <[email protected] <mailto:[email protected]>> writes:
We are eligible to apply for small grants for PETSc development under this program. Only one application is allowed per project. It can pay for access to resources (e.g. LLMs) or people's time with a 1099, so long as they are not already funded for similar work.
Ideas? I am thinking of one focused on PETSc LLM agents for coding etc.
Barry
Begin forwarded message:
From: Arliss Collins <[email protected]> Subject: [Projects] CFP - 2025 Round 2 Small Development Grants CFP is open Date: September 19, 2025 at 3:47:59 PM EDT To: [email protected] Reply-To: [email protected] Message-Id: <1758311278965.4a3368eb-b19f-40a1-a3ed-9043ac8a75a0@bf08x.hubspotemail.net>
Hello, Projects!
NumFOCUS is pleased to invite proposals for our 2nd round of Small Development Grants for 2025!
Reminder: Projects can submit a new proposal or re-submit a past proposal. Apply Here: <https://urldefense.us/v3/__https://d4jbd104.na1.hubspotlinks.com/Ctc/OT*113/... >small-development-grant-proposals <https://urldefense.us/v3/__https://d4jbd104.na1.hubspotlinks.com/Ctc/OT*113/... >
Timeline: Deadline for Proposal Submission: October 17, 2025 Projects will be Notified: November 21, 2025 (or sooner) Available Funding: Maximum $72,000 for allocation in Round 2. Up to $10,000 maximum per project for 2025. Eligibility: Any NumFOCUS Fiscally Sponsored or Affiliated project may submit 1proposal on behalf of the project per grant cycle. If you wish to solicit proposal ideas from your project community, the project leaders must organize their own review process to select the proposal that the leadership will put forward to NumFOCUS. Proposed work must be achievable within the 12 months following the award. The call is open to applicants from any nationality and can be performed at any university, institute, or business worldwide (US export laws permitting). Funding Distribution: For fiscally sponsored projects only, NumFOCUS can award the fund to the project in our accounting platform. The project would then submit invoices through the normal channels to pay the grant expenses. NumFOCUS can set up an independent contractor agreement, a must for contractor time over $600 USD, with an individual or company. You can see past successful proposals here: https://urldefense.us/v3/__https://numfocus.org/programs/small-development-g... <https://urldefense.us/v3/__https://d4jbd104.na1.hubspotlinks.com/Ctc/OT*113/... >
Please feel free to reach out <mailto:[email protected]> if you have any questions.
NumFOCUS, P.O. Box 90596, Austin, Texas 78709 Unsubscribe <https://urldefense.us/v3/__https://hs-39929437.s.hubspotemail.net/hs/prefere... > Manage preferences <https://urldefense.us/v3/__https://hs-39929437.s.hubspotemail.net/hs/prefere... >