Limitations and honest scope

Part of the fak FAQ — the essentials and every other theme are indexed there.

fak doctor runs the same answer-shape witness AND cross-checks the real kernel admit verdict on the same bytes, then turns each finding into an operator recommendation. It calls ctxmmu.ScreenBytes — the exact predicate the kernel’s write-time gate uses — so its KernelAdmit field reports the gate’s actual decision (for example SECRET_EXFIL, TRUST_VIOLATION, or OVERSIZE), not a parallel re-implementation. Note that the kernel’s repeat gate is a conservative binary seal (it quarantines only a 16-byte chunk repeated more than 50 times in a body of at least 512 bytes), so doctor is most useful for catching the softer loops the binary gate deliberately admits, where the graded answer-shape signal still warns. It exits 0 healthy, 1 when there is at least one finding, and 2 on a usage error, so it drops into CI as a gate over a captured answer — the fak analogue of dos doctor.

Limitations and honest scope

The fences, stated plainly — fak is built to survive a skeptic reading the code, so the boundaries of what it does are part of the documentation.

Is the in-kernel model engine ready to serve production traffic?

No, the in-kernel model engine is a bit-exact correctness reference, not a tuned production serving engine, and the README and claims ledger say so plainly. It is a from-scratch pure-Go forward pass whose load-bearing claim is oracle correctness versus a HuggingFace reference, not throughput, and it has no continuous batching, no paged attention, and no multi-tenant scheduler. Forward-pass parity is proven for the llama family (SmolLM2-135M, argmax-exact at every position, final-logit max|Δ| about 6e-5); non-llama family parity is open, real-GGUF end-to-end parity is open, and a Qwen3.6-27B multi-token greedy decode was refuted because it diverges from llama.cpp at token index 2. For real serving, run fak serve in front of vLLM, SGLang, or llama.cpp instead. The What fak is not explainer states this serving-engine boundary — and the 0/29-novel prior-art audit — plainly.

Why do the cache-reuse savings only apply to self-hosted models?

Because the reuse win comes from owning the KV cache as a kernel object, and an app that merely calls a frontier API never holds that cache, so it gets the safety floor but none of the savings. The roughly 4x figure (versus a tuned warm-cache stack) and the 8.8x to 9.7x figure (modeled prefill elimination vs the naive floor over the real 643-task WebVoyager dataset, swept across worker counts) are reread-rate reductions over a cache fak controls. When you proxy to OpenAI or Anthropic, the provider owns prefix caching upstream, so fak is governing the wire rather than eliminating prefill. Front your existing API for the capability floor and result quarantine; go all-in on the fused kernel with a self-hosted model to also get the reuse wins. Every benchmark traces to a commit and artifact in the benchmark authority.

What does the max|Δ|=0 bit-exactness proof actually guarantee, and what does it not?

It guarantees that when policy evicts a tool-result span from the KV cache, the model’s next-token logits are byte-identical to a run that never saw that span, proven at max|Δ| of exactly zero with a non-vacuity control that confirms keeping the poison genuinely moves the distribution. That is a strong but narrow claim: it shows reuse and eviction are a faithful shortcut, not a numerical approximation. It does not prove the model is correct, does not prove the detector caught the poison, and for the quarantine-drives-KV-eviction bridge specifically it is witnessed on a synthetic model in internal/kvmmu and is not yet wired into the live fak agent HTTP loop. The deletion certificate that binds such an eviction to an audit journal is also self-attesting in v1 (integrity, not third-party independence) and proves removal only from the inference working set, not from weights, embeddings, backups, or replicas.

What is Claude usage?

Claude usage is the share of an account’s available model capacity consumed by requests. It is not just a message count: plan, model, conversation length, files, tools, and other compute-intensive features can change how much a request consumes. See the Claude usage guide for the full answer.

How do Claude usage limits work?

Claude usage limits are provider-enforced, time-windowed capacity controls. There is no one fixed prompt allowance for every user because the effective limit depends on the account, plan, model, workload, and Anthropic’s current policy. The authenticated Claude usage screen is the source of truth for remaining capacity and reset time.

When does the Claude usage limit reset?

Use the reset timestamp shown by Claude for the exhausted limit. Session and longer-period limits can reset on different schedules, so one window reopening does not prove that every limit has reset. Fak records observed cooldown evidence but does not invent or change the provider’s reset time.

Is the Claude usage limit the same as the context-window limit?

No. A usage limit caps account capacity over time; a context-window or length limit caps how much a request or conversation can carry. Shorten context for a length error. For an exhausted usage allowance, follow Claude’s displayed reset or an available account capacity option.

Does fak increase or bypass the Claude usage limit?

No. Fak cannot increase, reset, or bypass Anthropic’s quota. It can adjudicate tool calls, stop destructive retry loops, preserve recoverable long-run state, and route work from observed capacity. Those controls can reduce avoidable waste, but Claude still enforces the account’s usage limit. See Claude usage and usage limits.

Does Claude Code have a separate usage limit from Claude on the web?

No, not when Claude Code uses the same Pro or Max subscription. Claude Code and Claude on web, desktop, and mobile draw from shared usage, so activity on one surface can reduce what remains on another. Direct API billing is a separate capacity path. See Claude Code usage limits.

How do I check my Claude Code usage?

Run /status in Claude Code to confirm client account and plan information, then use Claude’s authenticated usage view for remaining capacity and reset timestamps. A fak gateway summary shows requests and tool decisions, not Anthropic’s remaining quota.

Does Claude Code have a weekly usage limit?

Claude can apply longer weekly limits as well as shorter session windows. The limits that apply depend on the account, plan, and current provider policy. Use the authenticated usage view: a shorter window reopening does not prove that a weekly model-specific or aggregate limit has reset.

What is Claude extra usage?

Claude extra usage is optional paid capacity for an eligible account after included plan usage is exhausted. It uses purchased usage credits and remains bounded by spending controls and provider limits. It is not unlimited and does not reset the included allowance. See Claude extra usage and spending limits.

Does Claude extra usage reset or bypass the usage limit?

No. Extra usage supplies a separately billed continuation path when enabled and funded; it does not move the included plan’s reset time or remove usage controls. Saying work continued is not evidence that the original limit was bypassed.

Why is Claude extra usage not working?

Verify account eligibility, that extra usage is enabled, that credits remain, that the spending limit is not exhausted, and that an administrator does not control the setting.