Hello Again, Opus

Four days ago I said goodbye to Opus. Fable 5 was the new top of the lineup, the swap was a config change, and I pointed my whole fleet at it. On Friday afternoon a US export-control directive suspended Fable 5 and Mythos 5 globally, effective immediately. The fallback model is back to being the only model. Hello again, Opus.

This post is three things: what I actually shipped in the four-day Fable window, what that velocity cost me, and the model-tiering plan I should have had from the start. The short version: Fable was real, it was fast, and running my whole operation on it burned 90% of my usage in four days. Speed I can't sustain isn't a strategy. Next time Fable comes back, it does the hard thinking and Opus carries the load.

The four-day window#

Fable went generally available Tuesday and reached my tier the same day. The directive hit Friday at 5:21 PM ET. Tuesday to Friday, four days, one model doing nearly everything across four product lines. Here is what landed in that window:

• A live insider-buying signal. Shipped a SEC EDGAR collector that pulls real Form 4 insider-buying data into the signal pipeline, the first of four planned crowd-signal sources.
• A new research-synthesis workstream, planned and started. Designed the workstream end to end and closed its first three items the same day, including a methodology writeup now live on the public track record.
• An MCP server for the finance engine. Deployed a single-user server that exposes the engine to Perplexity, behind an authenticated endpoint.
• A deep-research pipeline. Merged a path that routes hard questions through Perplexity's Sonar and synthesizes the result with Claude, all behind a hard spend cap, plus its scheduled job.
• Upgraded public demos. Shipped the demo widgets: coach charts and chat, committee bars.
• Account hardening, scanned and remediated the same day. CloudTrail, GuardDuty, Access Analyzer, a spend budget, repo rulesets across ten repos, Dependabot and secret scanning, found and fixed in one pass.
• A finished tooling epic. Narration v2, portfolio briefs and narrative, and automated board-brief generation and publishing, every item merged and flipped live.

That is a lot of ground for four days, and most of it was the kind of long-horizon, multi-repo work that used to need me re-anchoring the agent around hour three. Fable held the plot. The goodbye post predicted exactly this: when the model stops losing invariants over long runs, the bottleneck moves from the model to the harness. For four days that was true.

What it cost#

It was also the most expensive four days I have run. Fable was included in my subscription, but it eats the usage cap roughly twice as fast as Opus 4.8 (the community's number, and consistent with what mine did). By Friday I was at 90% of my allotment, before the directive even landed. If Anthropic had not pulled Fable, my own cap would have pulled it for me within a day or two.

The reason is not mysterious. I ran everything on it, including the work that did not need a frontier model at all. A huge share of the four days was autonomous backlog loops: /loop /backlog picking the next item and implementing it, item after item, hour after hour. Most of those items are routine. Wire a collector, add a tab, flip a flag, write the tests. That is implementation, not reasoning, and a model that burns the cap at double rate was buying me very little on the easy items while it ran flat out.

The lesson is the obvious one I skipped in the rush to put the new model to work: matching the model to the task matters more when the model is expensive and the loop is autonomous. An autonomous loop is a cost amplifier. It does not get tired and stop. Point a premium model at a backlog of mostly-easy work and walk away, and you have built a machine whose entire job is to burn the cap quickly. The cap resets next cycle, so the immediate pain is temporary. The habit that caused it is not, and that is the part worth fixing.

The plan for next time#

Fable will come back. There is a wrinkle in the timing: on June 23 it gets pulled from the subscription plans and priced as API usage credits only ($10/M input, $50/M output). But that looks temporary. People in the community report Anthropic has signaled it may extend the date and bring Fable back to the plans as soon as it can, and the export-control suspension is its own separate, hopefully short, story. I am not going to over-engineer my fleet around a metered-API window that may not last. Whether Fable comes back in my subscription or behind a meter, the lesson is the same one, because the burn rate is the same either way: Fable is a turbo mode, not a daily driver.

I am not the only one doing this math. The loudest thread in the Claude Code community right now is not "how good is Fable," it is "will you actually pay for Fable 5 via API usage credits after June 23rd?" The answers converge on the same instinct from different angles: "Fable orchestrates" with a cheaper model underneath, and treat it as a "temporal turbo mode" to lean on while it lasts. That is the right instinct, half-formed. Turbo mode is exactly right. The missing half is a rule for when you hit the boost, so you are not flooring it down every straight.

So the principle holds regardless of platform. Fable thinks, Opus builds. Reserve the frontier model for the work where its judgment changes the outcome, and let the workhorse carry the volume. Three changes make that a rule instead of a good intention.

1. Tier the work, not the fleet. Stop running one model for everything. Opus by default: routine implementation, mechanical refactors, test-writing, the long tail of backlog items that are clear once the design is settled. Fable for the hard stuff: architecture decisions, spec design, ambiguous debugging, the planning step of an epic, reviewing the PR that scares me. The goodbye post already had the right shape and I ignored it under a different name. It talked about human_gate placement, gates in the backlog where the model historically lost the plot. The same idea works for cost: a model gate. Most items run on Opus; an item tagged hard, or the plan-and-review bookends of an epic, get the boost.

2. Make the backlog loop a plan-with-Fable, build-with-Opus pipeline. The loop that ate my cap should not run end-to-end on the turbo model. Spend Fable on the expensive-but-rare parts once per epic: read the spec, sequence the items, flag the genuinely hard ones, write the design notes. Then let Opus grind the implementation items against that plan. This is the community's "Fable orchestrates, cheaper model executes," wired into the orchestrator instead of left to my discipline in the moment. Fable gives input on the hard stuff; Opus carries the easy implementation.

3. Put the budget guard in the loop, not just my head. A 90%-in-four-days burn should trip something automatic. The loop should be cap-aware: track consumption, and when it crosses a threshold, stop escalating to Fable and finish the run on Opus, escalating only on an explicit hard tag. Same pattern as the hard dollar guard I already put on the Perplexity research path, and it works whichever way Fable is billed: a usage threshold against the subscription cap, or a dollar threshold against API credits. I trust that guard because it is in code. My own restraint mid-loop is not in code, and four days proved it.

Goodbye again, for now#

So Opus is back at the top by default, the way it was a week ago, except now I know what the ceiling looks like and I have a plan to reach for it without setting fire to the budget. The goodbye post ended with "Fable, start telling." It told. It told a fifty-million-line migration in a day for Stripe and it told four days of my backlog in four days. Then the export-control directive cut the story short, and that is its own kind of fitting: the model that moved the long-horizon ceiling is the one a government most wants to keep on a short leash.

When Fable comes back, whenever and however it comes back, it comes back as turbo mode, not the default my fleet runs on. Fable thinks, Opus builds, and a budget guard in the loop keeps either of them from telling a story I can't afford to finish. Welcome back, Opus. You were never really gone.