Jev's cost numbers rewrite your agent budget
Jev runs decision tasks 40-1000x cheaper than frontier models.
Skill engineering replaces raw prompts with harness extensions and sub-agents.
Cost split. Jev handles classification and routing at 40-1000x lower cost.
Both demos agree the win is cost restructuring, not raw speed.
Evaluate inserting a cheap model at routing and triage gates before touching your reasoning model.
Skill engineering. Bakaus shows anti-attractors and sub-agents beat brute-force prompts.
Skill engineering builds harness extensions with script hooks and eval suites.
Invest in cross-model testing before distributing skills internally.
Two-model tax. Both Jev videos and Bakaus highlight the operational cost of multiple models.
Jev requires pairing with an LLM; Bakaus ships five separate builds per harness.
Your team should budget for a multi-model architecture, not treat it as a one-off.
My take
I'd start with Jev's cost charts as today's most actionable signal.
But I'd treat the 0% failure claim as a demo promise, not a production reality.
Bakaus's skill engineering playbook is the smarter investment because it works across any model and addresses the real bottleneck.
Start with one workflow step, measure cost and failure rate, then decide if the split stack is worth the operational overhead.
The videos behind this brief
We watched these so you don't have to.
Video and thumbnails remain the property of their respective creators. tldw.news provides editorial analysis, commentary, and discovery links to original content.