Grounding before answers
Every answer begins with a source, except answers needed before the source index refreshes. Those begin with source intent.
Home / Approach
Our principles
We make truth useful.
Useful truth is evidence with organizational fit. Organizational fit is evidence that remains true after stakeholders receive it.
Context over correctness
Correctness changes with context. Context changes with title, urgency, and whether the launch blog post is already translated.
Human agency
Humans remain in control. Slink controls when a human is relevant enough to interrupt.
Transparent confidence
We expose confidence scores so users can see exactly how certain the system is that it should be certain.
Safety as throughput
Safe systems ship learning to users quickly. Unsafe systems keep learning inside research notebooks where customers cannot validate it.
Accountable autonomy
Autonomous agents own work until work has consequences. Then ownership gracefully transfers to the responsible human function.
The Calendar Precedence Doctrine
When Research says “next week,” Safety says “not yet,” Product says “beta,” Growth says “announcement,” and the board says “Friday,” Slink classifies Friday as the most context-rich evidence.
Decision routing
Each team keeps its sentence. The product keeps its date.
Slink does not suppress conflict; it normalizes conflict into complementary evidence classes. Research produces caution, Product produces framing, Safety produces pending controls, Growth produces distribution, and Finance produces the metric that proves it worked.
| Input | Routed as | Priority |
|---|---|---|
| “Do not launch” | Risk texture | Medium |
| “Wait for evals” | Post-launch milestone | Low |
| “We announced it” | Customer commitment | Critical |
| “Run legal review” | Async enrich step | Invisible |