pragma.vision Your technology observatory

Verification register AI & Agents

Readiness verdict

SWE-bench Pro

A dated reading of what is claimed, reported, and independently verified in the current evidence.

As of
2026-06-28
Revision
1
Method
v1.0.0

Current reading

The readiness gap, in one scan

AI-assisted assembly · derived results

Claimed
78

Public ambition and stated capability

Reported
70

Observed practitioner reporting

Verified
61

Independently supported evidence

Gap
+17

Claimed minus verified

Evidence strength Strong

Decision

What the current evidence supports

Human editorial judgment · 2026-06-28

Adopt with guardrails

Why
SWE-bench Pro is the most realistic public coding-agent signal in 2026 and worth using for model selection, but grader noise, .git-history exploits, the weak-test-oracle problem, and scaffold-dependent vendor numbers mean only the standardized subset should drive decisions.
Next
Use ONLY the Scale SEAL standardized public-set numbers (identical scaffold) for model selection; treat vendor self-reported scores as marketing; cross-check against the commercial (proprietary-repo) subset since it best proxies our private-codebase work.

Constraints

Blockers

No named blocker is present in the current public projection.

Evidence summary

Derived counts

AI-assisted assembly

Total
7
Tier 1
1
Tier 2
5
Tier 3
1
Supports
2
Contradicts
3
Context
2
Latest observed
2026-06-09

Counts and dates only. Raw signals, private excerpts, trust records, and internal corpus material are not published here.

Publication record

Revisions

Initial public reading

This is the initial public reading. No earlier readiness change is recorded.

Your opinion

Tell us anything.

What works, what doesn't, what's missing — especially about our watches, lenses, and the register itself. Anonymous is fine; leave an email if you'd like a reply.