The capability gap stated plainly…
The capability gap stated plainly: no model is optimized for finishing a task, and delegation needs exactly that.
The clearest statement of the Oddkin technical thesis: LLMs optimize for plausibility, delegation needs completion.
Transcribed from the asset above and checked against the record. Figures are reproduced as printed, including where the original is itself a claim rather than an audited number.
The Task Gap — no existing model is optimized for: verifiable task completion; persistent responsibility; failure detection and repair; outcome-based learning; escalation under uncertainty.
Why This Matters To Humans — real-world delegation will require models that: track commitments over time; act in environments with uncertainty; detect when plans fail; repair or escalate appropriately; optimize for completion, not plausibility.
LLMs optimize for token likelihood, reward models, preference scores, not verifiable task outcomes.
- Source
- project_Oddkin →
- File
- /images/notion/07d50470f00f4780.png
- Description
- Two-column slide contrasting “The Task Gap” with “Why This Matters To Humans”, over an orange banner about model optimization