What the internet is saying about GitHub Copilot

What the internet is saying about "GitHub Copilot"

Caveat: Direct discussion of Copilot is thin in this evidence. Most of the high-engagement items are about GitHub the platform (outages, growth) or AI coding tools generally. Where a point is adjacent rather than Copilot-specific, I say so.

🟢 Supporters

• •New multi-model product: GitHub announced "Copilot Fusion" to "route and orchestrate models" [1]. A related HN post, "GitHub Copilot Fusion orchestrates between models at half the cost," is the only cost claim in the data [3]. It is just a headline, and the evidence gives no methodology or numbers. Engagement was small (3 and 6 points).
• •Third-party tools are building on Copilot:
  • •GoModel added a github-copilot provider, authenticated with a GitHub OAuth token against api.githubcopilot.com and billed to a Copilot plan [8].
  • •The gh-aw-firewall project is working on Copilot subagent resilience [6] and provider-scoped model selection, such as "Claude over Copilot inference" [7].
• •Billing clarity: A toki2 fix prices Copilot Next Edit Suggestions at zero. A user's 374 NES requests had shown as "unknown cost," and the PR cites GitHub's billing docs on code completions [5].
• •Onboarding: GitHub's Skills "Getting Started with GitHub Copilot" exercise is in active use [2].

🔴 Skeptics

• •Integration friction (from tooling PRs):
  • •Copilot subagent requests "could fail after successful egress" when the selected model rejected the wire API or custom tools couldn't be translated [6].
  • •Unconstrained auto model selection "can select a model incompatible with the client's protocol" [7].
• •Cost surprise (Responses API budget steering): One reported run's warning request "cost ~9× a normal step and exhausted the budget before the agent could act" [9]. The visible text doesn't name Copilot specifically.
• •Platform reliability (adjacent):
  • •Repeated "GitHub Is Down" threads [20][30].
  • •"Another Friday, another GitHub incident" [24].
  • •A Pages/Actions/Webhooks degradation thread with 134 comments, where one commenter says they moved all CI/CD to Actions and that GitHub has "degraded measurably" [21].
  • •Roku moving its developer repos to GitLab, which drew 226 comments [19].
• •Security (adjacent): The Gheist campaign plants secret-dumping workflows using stolen GitHub tokens [17].
• •AI code review burden (adjacent): Reddit asks whether reviewing code now takes longer than writing it [10]. Another thread cites developers "tired of reviewing code nobody actually wrote" [13]. Neither names Copilot.

⚪ Neutral Observers

• •Scale: GitHub reportedly saw 7.4B commits in September and added 4M CPU cores and 160 PB of storage [11]. One commenter speculates that LLM-generated repos feed model training [11].
• •Security tooling: GitHub launched a package firewall in public preview [35].
• •Prediction markets: Polymarket has markets on GitHub critical-incident length and timing [34][36]. They are essentially illiquid, with liquidity of about $436 and $92 and zero volume, so they carry little signal.
• •Competitor chatter: Reddit discussion of Codex frustration [14], Cursor dropping third-party models [15], and frontier model weaknesses [16] doesn't mention Copilot in the visible text.

Cross-Platform Signals

• •GitHub is the most Copilot-specific source. It is mostly developer plumbing: provider integrations, model routing and billing fixes.
• •Hacker News is dominated by platform stability and trust (outages, Roku's move, security). Copilot Fusion got little traction there (3 and 6 points) compared with outage threads (159 and 310 points).
• •Reddit is mostly about AI coding workflows broadly, with no substantive Copilot opinion in this set.
• •Polymarket only reflects GitHub incident risk.
• •Consensus: There is no clear consensus on Copilot itself. The divergence is between active ecosystem building around it and anxiety about the underlying GitHub platform.

Community Verdict

Copilot is currently treated more as infrastructure than as a product to debate. Developers are integrating it as a model-access backend, and the Fusion multi-model routing announcement is the main news. Sentiment around it is shaped by GitHub's reliability problems and by wider fatigue over reviewing AI-generated code, not by direct opinions on Copilot's quality. Evidence on Copilot's own merits is sparse, so any firm verdict would be overreach.

---

Based on Hacker News + GitHub + Reddit + Polymarket data from the last 30 days, researched with last30days by Matt Van Horn.