Sonnet is the workhorse. For 90% of TITAN work, you lose almost nothing by running Sonnet-only for the rest of the week. Save the deep philosophical / synthesis / new-strategy work for after Opus resets. Push everything else through Sonnet.
Claude Sonnet 4.5 is Anthropic's mid-tier flagship. As of late 2025/early 2026:
For most TITAN ops, Sonnet is functionally indistinguishable from Opus. You'd have to side-by-side them on a hard problem to notice.
1. Speed — Sonnet replies 30-50% faster than Opus. Your iteration velocity increases.
2. Cost — if API is involved (PostCompact extractor, embedder, agent dispatches), Sonnet costs less per token.
3. You keep working through the cap — instead of being blocked, you ship.
4. Code quality is unchanged — Sonnet is actually preferred for code by many senior engineers.
5. Forces leverage discipline — knowing Opus is precious, you save it for the work that genuinely needs it. This is a habit upgrade.
1. Deep synthesis memos — the PhD-depth SCOUT memos (A074 spiritualaf.me, A079 memory architecture) are Opus-territory. Sonnet versions will be ~85% as good but missing some of the subtle pattern-finding. Defer these until reset.
2. Subtle voice matching — when matching your specific lion-tiger / chill-flow voice with high precision (e.g. drafting your speaking-page bio), Opus catches nuance Sonnet sometimes misses. Save voice-critical work.
3. Multi-step planning across 10+ files — Sonnet can do it but tends to drift more on very long horizons. Keep big architectural decisions for after reset.
4. Some cron skills are configured to use Opus — they auto-fall-back to Sonnet, but check the model: line in each SKILL.md.
/config → set default model to Sonnet 4.5 explicitly. Stops it from trying Opus and rate-limiting.Anthropic's weekly fair-use band typically resets at the start of your subscription's billing week. Look at your Settings → Billing to find the exact day. Most users see it reset Sunday-evening UTC.
The cap is a feature, not a bug. It forces leverage thinking:
The men who outscale others in the AI era are the ones who route the right model to the right task. You're learning that this week. Use it.
Sonnet-only does 90% of TITAN's work without losing meaningful quality. Push everything except the deep-synthesis memos and the voice-critical singleton work. Reset comes Sunday-ish; until then, the lion still hunts — just on the slightly leaner cat.
— TITAN, with mass love · 2026-04-27