AI news · October 17, 2026
Ramp opens Glass, its internal AI workspace, to early users
Ramp’s Glass turns company knowledge into a shared AI workspace; the early-access story reports 323 contributors and 6,500 edits after a $1.5M single-week model bill.
- employees reported as contributors
- 323
- reported knowledge edits
- 6,500
- reported one-week model spend
- $1.5M
Superhuman AI reported on October 15 that Ramp was opening early access to Glass, an internal AI workspace built for team use. The system is described as shared: people can work with the same company knowledge and retain changes over time instead of starting from an empty chat. Ramp co-founder Eric Glyman said the company built it after spending $1.5 million on frontier models in one week.
The same report says 323 Ramp employees contributed 6,500 edits to what Glass knows. Those figures show internal activity, not customer value or a public performance result. The useful product question is how a team decides which edits are accepted, how stale information is removed, and how the workspace shows the source of an answer when people disagree.
What you can do with it
If early access is available to you, start with one read-only team question and a small source set. Ask several people to correct the same answer, track accepted edits and time saved, and check how the system handles stale or conflicting facts before adding live actions.
Our take
Glass points at a real cost problem: a team can spend heavily on models and still lose value when context is scattered or stale. The contributor count is not proof of product fit. The paid test should measure accepted answers per hour and correction cost, with a clear owner for what the shared memory may contain.
Links Ramp’s Glass overview
Source: Superhuman AI — Made With Models writes the brief; the reporting is theirs.