And at the individual level, the 10x promise was real. Engineers are shipping more code than ever, and many are wiring up their own workflows to ship even more.
Record Acceleration
Pull requests merged90M
2023202420252026
Commits1.4B
2023202420252026
New repos per month20M
2023202420252026
source: GithubGitHub says agentic development workflows have sharply accelerated pull requests, commits, repository creation, API usage, automation, and large-repo workloads, forcing GitHub to move from planning for 10x capacity to designing for 30x scale.
Because individual productivity ≠ organizational productivity. Agents made individuals faster in silos, with no shared patterns and no shared memory or knowledge base.
10x productivity engineers
Chaos
AI organizations
Coordinated
We have more Pull requests10Drafts1111 Open48 Closed[WIP] feat(memory): cross-agent context sharing scaffold#622 opened 30m ago by sylg · Draft[WIP] refactor(orchestrator): pluggable agent backends#613 opened 2h ago by molishashah · Draftfeat(intent): spec diff viewer with conflict markers#609 opened 5h ago by marcus · Draftperf(retrieval): hybrid BM25 + dense rerank prototype#607 opened 12h ago by ravi · Draft[WIP] feat(workflow): Linear → branch automation#604 opened 1d ago by alex.k · Draftfeat(billing): grandfathered plan migration script#596 opened 2d ago by edinabazi · Draft[WIP] refactor(auth): JWT rotation policy#591 opened 3d ago by brianzye · Draft[draft] feat(slack): conversation memory across threads#588 opened 4d ago by maya · Draft[WIP] feat(cli): plugin installation flow#583 opened 1w ago by chase-of-the-fjords · Draft[draft] perf(builds): incremental compilation cache#580 opened 1w ago by divyapalaparthi · Draft[WIP] feat(monitoring): per-agent traces in Sentry#575 opened 1w ago by rinat-sharipov · DraftAwaiting your review4747 Open312 Closedfix(auth): clear stale session tokens on logout flow#624 opened 1h ago by marcusfeat(billing): annual plan proration on plan change#621 opened 3h ago by alex.kperf(api): add prepared statement caching for /search#620 opened 5h ago by ravifeat(slack): repo selector menu with type-ahead#619 opened 8h ago by mayafix(checkout): handle 3DS challenge timeout edge case#616 opened 1d ago by edinabazichore(deps): bump @sentry/node to 8.42.1#614 opened 1d ago by dependabot · Draftrefactor(ingest): split worker into producer/consumer#612 opened 2d ago by molishashahtest(intent): integration coverage for stale specs#611 opened 3d ago by sylgfix(blog): MDX code blocks losing language hint#608 opened 4d ago by rinat-sharipovfeat(cli): add --json flag to status command#605 opened 5d ago by brianzyedocs(api): regenerate OpenAPI spec from latest schema#602 opened 6d ago by osr-augment · Draftfeat(notifications): in-app toast for failed deploys#601 opened 1w ago by chase-of-the-fjordsfix(rate-limit): correct burst window for /v2 routes#600 opened 1w ago by divyapalaparthiPull requests1010 Open594 Closedfeat(pricing): add Cosmos NEW badge with hover tooltip to Max and Enterprise#623 opened 4h ago by brianzye · DraftLinkedin cards#618 opened 2w ago by divyapalaparthifeat: AI Transformation Assessment (Poseidon AI Consultant)#617 opened 2w ago by molishashah · Draftfeat(intent): redesign Intent product page#615 opened 2w ago by sylgAdd Intent SDD landing page concepts#599 opened Mar 25 ago by molishashahAdd RSS 2.0 feed for changelogs with client filtering#598 opened Mar 25 ago by osr-augment · DraftAi transformation#597 opened Mar 24 ago by divyapalaparthi · DraftMove CTA tracking, live session CTA, and experiments to guides/tools/learn#592 opened Mar 13 ago by edinabaziAdd Agent Readiness page with interactive framework#577 opened Mar 5 ago by rinat-sharipovfeat: Homepage v2 redesign with poster-style hero#572 opened Feb 28 ago by sylgfeat: Add Intent Community page with templates and showcase#567 opened Feb 25 ago by molishashahAdd new guides/tools/learn redirects#547 opened Feb 11 ago by edinabaziImplement Sanity Content Fields for SEO & PostHog Tracking#534 opened Jan 30 ago by edinabaziMoved the whole website into a layout group#531 opened Jan 30 ago by chase-of-the-fjordsAdd agentic loop with tool calling for web agent#530 opened Jan 29 ago by sylg, more tickets, and more alerts at 2am2:17Saturday, April 302:17PagerDutynowP1payments-api 5xx spikeError rate at 47% for 3 min · Paging primaryPagerDuty1 min agoP1queue-worker degradedworker_latency_p99 > 5s for 5 minPagerDuty3 min agoP2auth-svc memory pressureHeap usage at 92% of limitPagerDuty5 min agoP2checkout-flow timeoutsp99 latency > 8s · Customer reports risingPagerDuty8 min agoESCAuto-escalated to Tier 2No ack received within 5 min window than ever.
But nothing is compounding. Why?
A single engineer running a few agents in one client is not an operating model: it doesn’t connect the SDLC, it doesn’t turn one breakthrough into team knowledge, and it doesn’t teach the org what it learned.
That future needs a system, and we’ve been building it.
Watch: a 2-minute tour of Cosmos in action.
This is Cosmos.
The operating system for running agents across your software development lifecycle. Where humans, agents, code, tools, policy, and memory are coordinated at the org level.
It works where the team works, in the CLI, on the web, and on mobile, with the same build, tests, review, deploys, and conventions following you across all of them.
Every new agent inherits the context the team has already built up, so knowledge compounds instead of resetting with every session.
And when an agent gets something wrong, you don’t just fix the task, you teach the system: learnings from one incident carry forward, so the next run finds the next incident faster.
Agentscoordinate.remember.specialize.get better with the team.