Frontier RTL handoff study: replication and co-design call

[shsridhar-beep]

SV-Gap is opening a research-collaboration track around one question: when AI-generated RTL passes its functional evaluation, which production-review questions are still unanswered?

The primary next step is a 30-minute research scoping call. No code, internal RTL, private checkpoint name, or commitment to contribute is required. Start with one public, sanitized question; scheduling is coordinated separately.

The call should end with:

• a falsifiable research question;
• the smallest public or synthetic artifact that represents it;
• a claim and evidence boundary; and
• a go, revise, or out-of-scope decision for the next experiment.

Applied teams can continue through a bounded design-partner workflow: gap map, one-page experiment contract, minimal reproducible run, and evidence/continuation decision. Useful outputs include a model profile, diagnosis or repair study, public task, benchmark adapter, independent checker comparison, or documented negative result.

Start here:

• Request a research scoping call
• Review the design-partner workflow
• Run one packaged model task
• Browse public evidence profiles

Named contributor credit follows accepted, attributable work such as protocol design, redistributable model evidence, task design, analysis, validation, documentation, or code. Joining a call or providing an affiliation is not contributor status. Paper authorship is decided separately under the target venue's policy.

Do not post proprietary RTL, credentials, confidential constraints, private checkpoint mappings, or unapproved tool output. If you reached SV-Gap through a direct invitation, you may reply in that private channel with “SV-Gap research call” and a one-sentence question instead of opening a public issue.