Does GitHub still merit "top git platform for AI-native development" status?

The Pragmatic Engineer · email · 2026-04-03 source ↗ Reader 🖥️

Coding Agents Developer Experience & Flow Multi-Agent Orchestration Stacked Diffs / Stacked Pull Requests

GitHub's reliability has collapsed to roughly one-nine availability under AI-coding-agent load, a startup (Pierre Computer) is already demonstrating the AI-native git infrastructure GitHub lacks, and the root cause isn't primarily a scaling problem but an organizational one — GitHub has had no CEO or North Star strategy since Microsoft folded it into its AI group, leaving it chasing Copilot revenue instead of building the agent-native platform its own infrastructure crisis shows it badly needs.

practical · mixed — A newsletter analysis piece (Gergely Orosz, re-sent as a free bonus issue; the paying-subscriber original ran about eight days earlier) mixing reported outage data, a quoted third-party technical analysis (Lori Hochstein), a quoted practitioner prescription (Mitchell Hashimoto), and the author's own diagnosis of GitHub's organizational dysfunction.

Structure

Part	Role
Data: GitHub's availability has dropped to ~one nine (~90%), tracked by a third-party status page after GitHub stopped updating its own; Claude Code's share of GitHub load has grown 6x in three months	evidence
The three named incidents (2 Feb, 9 Feb, 5 March) and Lori Hochstein's technical analysis of them as a cycle of infra strain compounding failovers	evidence
Startup contrast: Pierre Computer's Code.storage claims throughput (>15,000 repos/min peak vs. GitHub's ~230/min) GitHub cannot match	evidence
Mitchell Hashimoto's quoted prescription: a North Star around agentic-code infrastructure, killing Copilot, acquiring Pierre, and re-evaluating all product lines against that North Star	evidence/alternative framing
The author's own diagnosis: Copilot/Microsoft internal politics, no CEO since Thomas Dohmke's departure, and resulting lack of focus (chasing Copilot revenue, shipping stacked diffs that AI agents may make obsolete)	conclusion

Key terms

North Star (for GitHub)
Hashimoto's term for a single, measurable strategic anchor — specifically, being critical infrastructure for agentic code lifecycles — against which every GitHub product line and initiative should be re-evaluated, with the implication that most currently fail the test.
agent mailboxes
Hashimoto's proposed first-class GitHub platform primitive for agent-to-agent and agent-to-human messaging, distinct from GitHub itself acting as an agent — GitHub should be 'a platform and not an agent itself.'
AI-native git platform
The article's titular frame: a code-hosting platform architected from the ground up for the request volume, repo-creation rate, and traffic pattern AI coding agents generate — which the author argues GitHub currently is not, despite having an MCP server.

Propositions

explicitGitHub's availability has fallen to roughly one nine (~90%, about 2.5 hours of daily issues/degradation), down from the four-nines standard expected of reliable infrastructure.conf: high
explicitGitHub stopped updating its own status page, reportedly because of how poor its true availability had become, prompting a third party to build a replacement tracker.conf: medium
explicitLoad on GitHub from Claude Code's bot contributions alone grew roughly 6x over three months, a concrete proxy for the broader AI-agent-driven traffic increase GitHub's infrastructure is failing to absorb.conf: medium
explicitGitHub's named February and March 2026 outages share a recurring pattern — infrastructure strain triggering failovers, and failovers exposing configuration or telemetry gaps that make the incident worse than the original trigger.conf: medium
explicitDatabases are harder to scale up under sudden load than stateless services, and GitHub underestimated how much additional AI-agent traffic its database layer would need to absorb.conf: medium
explicitA startup (Pierre Computer, via its Code.storage product) has already demonstrated repo-creation throughput (self-reported: >15,000 repos/minute sustained peak, >9M repos in 30 days) that vastly exceeds GitHub's reported ~230 new repos/minute, suggesting GitHub's architecture is not competitive for AI-agent-scale git hosting even against a young competitor.conf: low
supportsThe Pulse: AI load breaks GitHub – why not other vendors? · p8 — Pierre Computer/Code.storage is a concrete, named instance of the market opening for new, AI-scale-architected code-hosting entrants that extract-ai-load-breaks-github predicted GitHub's crisis would create.
explicitIf GitHub kept failing to solve this reliability problem, teams would begin trying alternatives (startups like Pierre, or self-hosted Git) rather than remaining locked in by default.conf: medium
explicitGitHub should establish a single North Star around being critical infrastructure for agentic-code lifecycles, shut down Copilot, acquire Pierre to launch agentic repo hosting as its first agentic product, and cut roughly half its product lines/initiatives that don't serve that North Star.conf: medium
explicitAll agentic interactions with GitHub should rely on first-class platform primitives (agentic code review, an agent chat primitive, agent mailboxes) rather than being bolted onto the existing product via GitHub Actions — and GitHub itself should be a platform other agents build on, not an agent competing with them.conf: medium
supportsSteve Yegge's Vibe Coding Manifesto: Why Claude Code Isn't It & What Comes After the IDE · p6 — Hashimoto's 'agent mailboxes' platform-primitive proposal and Yegge's description of Jeffrey Emanuel's 'Agent Mail' tool are independent arrivals at the same idea — a messaging/mailbox layer for coordinating agents — one proposed as a GitHub platform feature, the other built ad hoc by a practitioner; both treat this as missing infrastructure for exactly the team-scale agent-orchestration problem Yegge's 'agent orchestration dashboard' answer is reaching for.
explicitGitHub's Copilot product and brand are entangled with Microsoft-wide internal politics — Microsoft extended the 'Copilot' brand across many products after GitHub Copilot's early success, while internal Microsoft AI/Azure groups separately sought control over GitHub as a valuable developer brand.conf: low
explicitGitHub has had no CEO since Thomas Dohmke's voluntary departure, and Microsoft deliberately did not backfill the role, instead folding GitHub into Microsoft's AI group and stripping its independence.conf: medium
extendsThe Pulse: AI load breaks GitHub – why not other vendors? · p4 — extract-ai-load-breaks-github attributes GitHub's crisis to self-inflicted technical causes (tech debt, org debt, Azure migration); this document adds a specific organizational mechanism behind that debt — no CEO since Dohmke's departure and Microsoft-internal Copilot politics — without which the 'organizational debt' claim stays abstract.
explicitLacking a CEO or clear direction, GitHub's teams default to the locally safest strategy — investing further in Copilot to grow revenue — rather than addressing longer-term structural issues like reliability.conf: low
explicitGitHub Copilot's position as the most-used AI coding agent has already been overtaken by Claude Code and will soon be overtaken by Cursor, evidence that GitHub's core AI product is losing its category leadership.conf: medium
explicitcontestedGitHub's work on stacked diffs (started October 2025) may be largely obsolete by the time it ships, because AI agents change the review workflow the feature was designed for.conf: low
contradictsGitHub Brings Stacked Pull Requests to Public Preview · p2 — extract-github-stacked-pull-requests frames stacked PRs as solving the human-code-review bottleneck; Hashimoto's claim (reported here) that the stacked-diffs workflow 'might be mostly obsolete' with AI agents challenges the premise that this is still the bottleneck stacked diffs need to solve, without itself substantiating why.
extendsSteve Yegge's Vibe Coding Manifesto: Why Claude Code Isn't It & What Comes After the IDE · p7 — Yegge names Graphite/stacked diffs as the best-poised (but not yet successful) attempt to solve multi-agent merge collision ('the wall'); this document's claim that GitHub's own stacked-diffs effort may already be obsolete by the time it ships adds a concrete, dated data point to that unresolved race between tooling and agent-driven change volume.
implicitA platform trying to do many things at once (Copilot, core git hosting, AI infra) while excelling at none is structurally weaker than a focused competitor doing one thing better than everyone else.conf: medium

Arguments

GitHub's AI-era crisis is fundamentally an organizational/leadership failure (no CEO, internal Microsoft politics, no North Star) rather than a purely technical scaling problem. strength: medium

Explicit premises:

GitHub has lacked a CEO since Dohmke's departure, with Microsoft declining to backfill the role.
GitHub Copilot and GitHub itself are entangled in Microsoft-internal brand and organizational politics.
Without a CEO or clear direction, teams default to the locally safe choice (Copilot revenue) over structural fixes (reliability).

Implicit premises:

A reliability/infrastructure crisis of this severity at a company with Microsoft's resources is better explained by organizational dysfunction than by resource scarcity or an intractable technical problem.
The absence of a formal CEO is itself sufficient to produce directionless teams, rather than other explanations (e.g. a deliberate matrix/committee structure that could in principle still produce direction).

Inferences:

If the diagnosis is organizational, then technical fixes alone (more database capacity, better failover testing) will not resolve the underlying pattern of neglect — only a leadership/structural change would.

GitHub should narrow to a single agentic-infrastructure North Star and treat its current reliability crisis as proof it needs exactly the kind of AI-native platform Pierre Computer is already building. strength: low