01 — geeky-gadgets.com
https://www.geeky-gadgets.com/claude-code-cost-optimization/
Editor's note — #video #cost-control #claude-code — Written companion to a Simon Scrapes video. .claudeignore files and loop engineering to cut token burn.
Research set on Anthropic's Mythos-class Fable 5 model. Covers when Fable earns its 2x price over Opus 4.8, the levers for controlling cost (effort, caching, batch, routing), and head-to-head comparisons. Tagged by topic and source type in each link's note.
01 — geeky-gadgets.com
Editor's note — #video #cost-control #claude-code — Written companion to a Simon Scrapes video. .claudeignore files and loop engineering to cut token burn.

02 — YouTube
Note — #video #fable-vs-opus #open-weight — AICodeKing. Adds GLM 5.2 as an open-weight comparison point.

03 — YouTube
Note — #video #fable-vs-opus #head-to-head — AI Foundations. Fable vs Opus 4.8 run side by side.

04 — YouTube
Note — #video #fable-vs-opus #head-to-head — Post-restoration head-to-head against Opus 4.8.

05 — YouTube
Note — #video #workflow #production — "Better Than 90% Of People." Production/team workflow orientation rather than a demo.

06 — YouTube
Note — #video #review #when-to-use — "Honest Review (updated)." Weekend of real project testing; strong on planning and spec work.

07 — YouTube
Note — #video #explainer #when-to-use — AI for Mortals, "No Hype Full Breakdown & Testing." Best straight explainer of the video set.

08 — The New Stack
Note — #fable-vs-opus #contrarian #cost-control #press — The New Stack. Identical tests: results converged, bills didn't.

09 — tessl.io
Note — #fable-vs-opus #contrarian #benchmark — Data-backed skeptical read. ~1,000 scenarios scored twice: Opus 4.8 still better value for most agent fleets. Read against TrueFoundry.

10 — alexmcfarland.substack.com
Note — #fable-vs-opus #decision-framework #effort — Sharpest single insight: "I need Fable" is often really "I need Opus actually trying." Ladder = Sonnet high / Opus xhigh / Fable by the job.

11 — truefoundry.com
Note — #fable-vs-opus #decision-framework #routing #benchmark — Cleanest decision framework of the set. Also covers gateway routing setup.

12 — Claude Fast
Note — #cost-control #pricing #subscription-vs-api — Credits vs straight API key accounting. Full 1M context at standard rates, no long-context surcharge.

13 — costlens.dev
Note — #cost-control #governance #budgeting #enterprise — The runaway-spend piece. Uber blew its 2026 AI budget by April, now caps engineers at $1,500/mo. Case for kill switches on agent loops.

14 — Medium
Note — #cost-control #aggressive #effort #benchmark — Claims 80% reduction. Fable at low effort (~$3.76/task) beating Opus at max (~$13). Some techniques are exotic — vet before adopting.

15 — buildthisnow.com
Note — #cost-control #pricing #effort — Five levers: effort, task budgets, caching, batch, routing. Good on the fallback pricing quirk.

16 — TokenMix
Note — #cost-control #pricing #batch #math — 7 levers with all arithmetic shown. Batch at $5/$25 (half price both sides) for anything async.
17 — upstash.com
Note — #cost-control #effort #caching #routing — START HERE for cost. Effort tuning, prompt caching (~1/10th input price), model tiering with subagents.

18 — Medium
Note — #use-cases #when-to-use #non-coding — Week of real use. Concrete deep-research and knowledge-work workflows, not just coding.

19 — lennysnewsletter.com
Note — #review #when-to-use #video #podcast — Lenny Rachitsky, early access. Where it fits in your AI stack. Article + YouTube/podcast versions.

20 — productcompass.pm
Note — #analysis #benchmark #when-to-use #effort — Best empirical piece. 7 experiments, 1,000+ timed runs. Publicly corrected two claims that failed retesting.

21 — Simon Willison’s Weblog
Note — #analysis #when-to-use #hands-on — Simon Willison day-one impressions. "Slow, expensive, and quite happily churning through everything."

22 — Amazon Web Services
Note — #official #deployment #enterprise #safeguards — Bedrock access path. Notes bio/chem/health/cyber prompts fall back to Opus 4.8 — relevant for regulated workloads.
23 — Claude Platform Docs
Note — #official #docs #prompting #when-to-use #effort — Most useful official page. Effort levels, long-run behavior, timeout/harness redesign. "Apply it to your hardest unsolved problems."
Subscribe to be emailed when “Claude Fable 5: When to Use, Cost Control, vs Opus” changes. Unsubscribe anytime.
Comments
Sign in to join the conversation.