Scores that understand your code.
Silk 1 is Weave’s new code output model. 3.3x more accurate, with plain-language reasoning for every score and calibration that learns from your team. Same scale. Same API. Better signal.
Proportion of code output generated by AI compared to all code output
Scroll horizontally to explore all weeks →
- Human output
- AI output
Average error vs expert estimate3.3×lower average prediction error · 88% of PRs land within ±1h
Who it’s for
Built for every layer of the engineering org
Silk 1 surfaces differently depending on where you sit.
Engineering managers
Accurate output data across your team without second-guessing the numbers. Automated reporting you can trust and share with leadership as-is.
VPs and CTOs
Board-ready data on engineering productivity powered by a model that understands your codebase. Make investment decisions based on signal, not estimates.
Engineers
Transparent scores with clear reasoning. Understand how your work is measured and flag scores that don’t look right. Your overrides make the model better for everyone.
What’s better
Silk 1 reads the full diff and reasons about intent, risk, and interdependencies across files.
- 3.3xReduction in average prediction error
- 88%Of PRs scored within ±1h of expert estimate
- 35xMore model capacity than the original
Old model
Sees 3 lines changed. Scores for line count.
Silk 1
Understands the downstream impact across 40 files. Scores for true complexity.
Example
A 3-line config change that unlocks a migration path across 40 files.
config/migrate.ymlstrategy: legacystrategy: incrementalparallel: true
What the model does
Four capabilities the old model didn’t have.
Deeper code comprehension
Understands what a code change does and why it’s complex. Analyzes the complexity of each line changed and reasons about cross-file dependencies.
Transparent reasoning
Every score ships with a plain-language explanation of how the model arrived at its estimate. When a score looks off, read the reasoning instead of guessing. No other engineering analytics tool offers this level of transparency.
Organization-aware calibration
Uses reinforcement learning to learn from your team’s score overrides. The model adapts to your codebase’s conventions and complexity profile over time. The longer you use it, the more accurate it gets.
Cross-file understanding
Reasons about how changes in one file affect others. Complex refactors and cross-cutting changes finally get the scores they deserve.
What stays the same
Zero migration required.
Silk 1 uses the same scoring scale, the same API, and the same report format. A “3.0” still means three hours of expert engineer effort.
- Same scoring scale
- Same API
- Same report format
- Dashboards, integrations, and alerts work without modification
Measure engineering output with the most accurate scoring model in the industry.
Get started in 5 minutes or book a demo with our team.