Large CSV exports no longer time out
Accounts above 100,000 rows now stream their export instead of failing partway through, resolving a recurring problem for large customers.
Engineering news, not engineering noise
Engwire finds the few engineering changes worth communicating, explains what changed for the people who need to know, and sends the update to the right Slack channel. No firehose. No status report to write.
For B2B software teams with 8–40 engineers using GitHub + Slack. We're onboarding the first six teams by hand.
Arrives in Slack. No dashboard to check.
Large CSV exports no longer time out
Accounts above 100,000 rows now stream their export instead of failing partway through, resolving a recurring problem for large customers.
8 activities → 1 update worth reading
What the tools record
Technically visible. Practically invisible — which is why the channel is muted.
What people need to know
Large CSV exports no longer time out
Accounts above 100,000 rows now stream their export instead of failing partway through.
One update, after production availability is proven. Not when the branch merges.
“Our GitHub channel is muted.”
“Someone writes the weekly engineering update by hand.”
“Support keeps asking whether it shipped yet.”
They're the same problem: engineering systems record what happened, but nobody turns it into a useful update for the people outside the workflow. Engwire automates the factual spine; humans keep the judgment.
Engwire recognizes state changes; it does not forward activity. Detection is deterministic — explicit rules over source evidence, with no model in the path.
Fires once when a pull request actually becomes reviewable.
draft removed · opened non-draft · reopened non-draft
Never on commits · passing checks · reviewer assignments · labels · comments · merge
Fires once when a delivery is proven available in production.
One deployment carrying twenty merged pull requests becomes one update — not twenty.
based on the availability boundary configured for that scope
merge ≠ shipped · staging ≠ shipped · green build ≠ shipped
Same evidence. Different audience.
Engineering
Streaming export path is live in production
Export requests above the row threshold now stream from the worker instead of buffering the full result set.
Support / CS
Large CSV exports no longer time out
Customers with very large accounts can export their data without the download failing partway through.
Readers don't need an Engwire account. The news arrives in Slack.
01 — Evidence
Source facts are retained and normalized. Duplicates and out-of-order delivery resolve safely.
02 — Detection
Explicit deterministic rules decide whether a meaningful event occurred. No model makes this decision.
03 — Routing
You state intent: shipped Billing changes go to #support-heads-up. Generated text cannot decide destinations.
04 — Wording
A model explains known facts at the level the audience needs. When evidence is thin the claim is omitted, not softened.
Every update keeps its provenance, so you can see why it fired and what produced its wording. If generation is disabled or fails, the update still goes out — built from evidence.
Transparency, not surveillance
Making outcomes visible must never become scoring people. This is a product decision, not a missing feature.
Six design partners, then we open it up
We'll connect GitHub and Slack with you, watch the first weeks of real updates, and tune anything that arrives as noise.
Who this fits
A short email, then a shared setup call. Every GitHub and Slack permission we ask for is listed before you authorize anything, and uninstalling deletes the stored activity.