Ornith-1.5: Self-Scaffolding to Self-Improvement

I'm Matt Wood, and this is For Your Information. A live list of riffs and links for you and your agent, drawn from what I'm reading, noticing, questioning, concluding, and revising.

Ornith-1.5 introduces an end-to-end self-improvement framework for foundation models that continuously generates new tasks, creates task-specific scaffolds, and produces solution rollouts for reinforcement learning. Available in three scales (397B, 35B, and 9B parameters), the model achieves state-of-the-art performance among open-source models on reasoning, coding, and agentic tasks, matching Claude Opus on key benchmarks.

ornith.ai →

Connections

Supports

The Economics of Recursive Self-Improvement Ornith-1.5's self-scaffolding and continuous task generation is a concrete technical implementation of the recursive self-improvement economics discussed in 554B2536

Related to

Inkling: Thinking Machines' Open-Weights Model (and Tinker Fine-Tuning Platform) Both involve open-weights models with fine-tuning/improvement platforms; Ornith-1.5's self-improvement pipeline relates to Inkling's Tinker fine-tuning approach
Building an Advanced Agentic Harness Both address advanced agentic harness construction; Ornith-1.5 auto-generates scaffolds while 3FA0663B focuses on manually building advanced agentic harnesses
SIA: Self Improving AI Framework Both are self-improvement AI frameworks; FF88FEE8 is a self-improving AI framework while Ornith-1.5 specifically focuses on end-to-end self-scaffolding to generate tasks and rollouts for RL

Develops into

Managing Agent Skills with Dr. Skill Ornith-1.5's task-specific scaffold generation is a more automated evolution of the skill management concept explored in Dr. Skill for managing agent capabilities

Type:	link
First published:	Wednesday, 19 August 2026 at 19:49 UTC
Location:	https://ornith.ai/ornith_1_5.html
URL:	https://mattwood.fyi/i/286593b8