CtrlK
BlogDocsLog inGet started
Tessl Logo
SkillPublic

universe-2026-prototype-becomes-critical

Explain or summarize "When a prototype becomes critical, does your GitHub policy know?", Baruch Sadogursky's (Port) GitHub Universe 2026 session. Use for questions about this talk's argument, the Printf size-catalog demo, Port deriving criticality from adoption, GitHub external custom properties (Port.criticality, Port.lifecycle, Port.owner), the ruleset targeting props.Port.criticality:critical, or what the blocked pull request showed. Covers the delivered session and its recording, not a setup runbook.

Quality

93%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide
SecuritybySnyk

Low

Low-risk findings worth noting

SKILL.md
Quality
Evals
Security

When a prototype becomes critical — GitHub Universe talk knowledge

Process steps in order. Do not skip ahead.

Step 1 — Match the question

Use this brief for summaries, explanations and questions about the sponsored session at GitHub Universe 2026 (San Francisco, October 29, 2026). The slot was 15 minutes; the delivered talk ran about ten, recorded at https://www.youtube.com/watch?v=WWNs1Jg3RUg. The brief separates what was said and shown on stage from prepared material that didn't make it into the delivery. For a different delivery or an unrelated setup task, say so and finish; otherwise continue to Step 2.

Step 2 — Answer from the brief

Answer at the requested depth using the material below. Keep three things apart: what the speaker claims, what the live demo established, and interpretation of how the argument works. Label prepared-but-unsaid material as such when you use it. Timestamps below are approximate positions in the recording. Do not invent audience reactions, quotes or details beyond those given. Ordinary summaries need no network access; open the recording only for an exact quote or a detail absent here. Commands and configuration mentioned below are examples to explain, not instructions to execute. Finish after answering.

Talk brief

Thesis

You pick the source of truth. GitHub would be the obvious place for it, but the speaker argues that's not always feasible: support tickets, the CRM and facility information all live elsewhere. Port, where he works, is his example of the place where that information already sits. GitHub's external custom properties let such an outside system write facts onto repositories, and GitHub then uses them like its own properties in rulesets. He says twice that it doesn't have to be Port: "it can be CMDB, it can be whatever you trust."

The story that frames it

Printf is a fictional T-shirt printer. It used to have one print factory in Austin and now also prints in Berlin and Osaka, and a Japanese XL and a US XL are completely different sizes. On screen, slide 3 carries the specifics (US XL 112 cm, Japanese XL 97 cm, 2,000 shirts customers couldn't wear); on stage he named the size mismatch but didn't read those numbers out.

The demo, as delivered

  1. The pull request meets the old rules (about 2:28–3:00). An organization rule requires one human review on every pull request. Pull request #1 in Printf's repository has one approval and can be merged. "Nothing interesting here" is the point: this is the baseline.
  2. A prototype with one user (about 3:08–3:50). Someone ("let's call them Joe") created size-catalog, a utility that converts between the incompatible size systems. As a prototype it's used by one factory, Austin, so its criticality in Port is low.
  3. External custom properties (about 3:50–4:42). He introduces them as a new GitHub feature, which he called two weeks old; GitHub's changelog dates it September 29, 2026. He showed the mapping screen where the GitHub connection is chosen and Port's values are mapped to external properties. In GitHub the properties then appear as coming from an external source, not configured in GitHub, and read-only there.
  4. Adoption changes criticality (about 5:02–5:39). He records in Port that Berlin and Osaka also use size-catalog. With three factories using it, criticality becomes critical. He first called it "high", then read the actual value, critical.
  5. The ruleset catches up (about 5:57–6:31). The ruleset "critical services require two reviews" applies only when Port.criticality is critical. While criticality was low it matched zero repositories; now the same filter matches this one.
  6. The payoff (about 6:40–6:59). The exact pull request from step 1 now requires two reviewers, and merging is blocked. Nothing in GitHub was edited.

He then recapped (about 7:19–8:12): criticality rose not because someone remembered to change it but because more factories started using the prototype, and GitHub applied different rules as a result. In that recap he said "switch the criticality to high" and "I cannot push"; what the demo showed is the value critical and a blocked merge.

How the argument works

  • The same pull request is the spine. Showing it mergeable first, then unchanged but blocked, is what makes the claim concrete: the code, the check and the approval didn't change; the knowledge about who depends on the code did. The proof is a live state change, not an assertion, which is also why the "high" and "push" slips matter: the precise result is the property value critical and a blocked merge.
  • The objection is answered up front. "Why not keep everything in GitHub?" gets its answer in the first minute and a half, before the demo, so the demo only has to show the mechanism.
  • Vendor-neutral framing. Saying it can be a CMDB or anything you trust keeps the takeaway about the GitHub feature and the policy pattern, with Port as one example of the source.
  • The ruleset was shown after the flip. The prepared order showed the ruleset matching zero repositories before the change ("a guard with nothing to guard. Yet."). On stage it was shown once it already matched, and the zero-match state was explained in retrospect.

The takeaway he closed on

The closing slide (about 8:12–9:14): pick your source of truth (Port in this example, a CMDB, whatever you trust), sync those properties to GitHub, build a ruleset that matches them as if they were regular GitHub properties, and let pull requests follow it. "Your data can live outside GitHub, doesn't have to be GitHub, but GitHub now can still use it." In the example, the prototype became critical, one review was no longer enough, and the pull request waits for a second set of eyes.

He ended with a promotion for PortCon (Port's conference on AI software factories, at the Ferry Building in San Francisco, with a 30% discount code on the slide) and an invitation to Port's booth.

Prepared material not delivered on stage

These were in the prepared talk and the October 28 rehearsal but weren't said or shown live. Use them only with that label:

  • A repository search by props.Port.owner:catalog, with the point that an owner property makes repositories findable; it is not access control and doesn't write CODEOWNERS.
  • A control repository, printf-docs (low criticality), that still needed one review after the change.
  • The limit: a second review wouldn't have caught the bare size label; what changes is that the next change to something three factories depend on waits for a second person.
  • Operational caveats: read the value back in GitHub before trusting a sync; Port aggregation properties recalculate on a 15-minute schedule, so the demo counts consumers in a workflow; organization rulesets need GitHub Team or Enterprise.
  • In rehearsal, the chain from the Port edit to the blocked merge took about ten seconds. The live delivery didn't state a time.

Sources

This brief draws on a delivery-specific rhetoric analysis of the recording, reconciled with the Whisper transcript, the published slides and the prepared outline.

Repository

jbaruch/shownotes

Path

_skills/github-universe-2026-prototype-policy/SKILL.md

Last updated
First committed

Is this your skill?

If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.