Ask HN: Anyone's Claude Code session finishing quickly from yesterday?

A Hacker News Ask HN thread raised a concrete usage complaint: on a $100 Claude Code plan, the 5 hour limit was burning down far faster than expected starting the day before the post. The reporter also had the clau.de/cc-50-promo promo enabled and tied the change in burn rate to a Claude status incident (status.claude.com/incidents/mgp99sn4ynd4). At the time of the summary the thread had 1 point and 3 comments, so this is early signal, not a broad survey.

Claude Code’s practical meter for heavy users is that 5 hour window on the $100 plan. When session budget collapses mid-work, the failure mode is not a hard product outage; it is interrupted agent loops, truncated tool runs, and forced pauses right when a coding session is most state-heavy. Linking the behavior to a named status incident implies users are treating platform reliability events as possible causes of metering or accounting anomalies, not only of raw availability.

For engineers who run long agentic sessions, quota is capacity planning. A limit that finishes early breaks batch refactors, multi-file fixes, and unattended coding loops that were budgeted against a known 5 hour envelope. Promo stacking (here, the cc-50-promo on top of the paid plan) makes the accounting harder to reason about: if effective usage rate or remaining allowance is unclear after an incident, builders cannot trust remaining session time as a stable resource.

In the Claude / Codex / Fable competitive set, coding agents compete on model quality and on predictable cost-of-use. A public Ask HN about sudden 5 hour exhaustion on a $100 plan lands as a trust issue: peers will compare whether Claude Code, Codex, or other tools give steadier session budgets under similar workloads. Sparse early engagement (1 point, 3 comments) still matters because power users watch these threads for shared symptoms after status events.

What to watch next is whether more $100-plan users report the same post-incident burn, and whether Anthropic’s incident notes or support clarify if metering, promo accounting, or session handling changed around that status event. Until then, treat remaining Claude Code session time as unverified after any related incident: checkpoint work often, avoid assuming the full 5 hours, and keep a fallback path (local tooling or another agent) when a session dies mid-task.