# Contested concept: agentic harness

_**What you are reading**: one contested-concept page in Deliberus's deliberation graph. Deliberus maps human reasoning — including claims that are contested, mistaken, or extreme — so that disagreement can be examined precisely. A claim's presence here is not an endorsement: the graph records that someone argued it, how it connects to other reasoning, and how far scrutiny has gotten. The metadata on this page describes the state of that scrutiny, not the truth of the claim._

---

- **Contestedness**: semantic
- **Diagnosis**: This semantic shift masks a disagreement about the future of AI engineering: is it a temporary 'gap-filling' exercise that disappears as models improve, or a permanent discipline of system architecture?
- **Canonical**: `/concept/agentic harness` (data: `/concept/agentic harness.json`)

## Senses in use
- **A structural combination of models and processes that creates new, higher-order capabilities regardless of the underlying model's baseline.** — used in 4 claim(s)
- **External scaffolding designed to compensate for specific technical deficiencies or 'limitations' of an underlying model.** — used in 4 claim(s)

## Claims using this term
- [As large language models improve, the focus of AI engineering shifts toward discovering new and novel combinations of [scaffolding and agen…](/claim/claim_34d51cfed617.md) _(sense: A structural combination of models and processes that creates new, higher-order capabilities regardless of the underlying model's baseline.)_
- [The primary and most valuable task for AI engineers [in the context of agentic development] is the continuous discovery of novel combinatio…](/claim/claim_406b4c9086bf.md) _(sense: A structural combination of models and processes that creates new, higher-order capabilities regardless of the underlying model's baseline.)_
- [In the development of long-running coding agents using Claude Opus 4.5, it was possible to remove 'context resets' (clearing the context wi…](/claim/claim_42ed3e59375d.md) _(sense: External scaffolding designed to compensate for specific technical deficiencies or 'limitations' of an underlying model.)_
- [Increases in the baseline capability of large language models create more opportunities to develop [agentic] harnesses capable of performin…](/claim/claim_47bdabd6b87d.md) _(sense: A structural combination of models and processes that creates new, higher-order capabilities regardless of the underlying model's baseline.)_
- [Assumptions about large language model limitations in harness design become obsolete as the capabilities of the underlying models improve o…](/claim/claim_54a76ab9ad79.md) _(sense: External scaffolding designed to compensate for specific technical deficiencies or 'limitations' of an underlying model.)_
- [The variety and utility of interesting [AI agent] harness combinations [structural scaffolding surrounding large language models] does not …](/claim/claim_7d35b1898fac.md) _(sense: A structural combination of models and processes that creates new, higher-order capabilities regardless of the underlying model's baseline.)_
- [AI engineers should proactively remove components from a model's harness that are no longer essential for the performance of the underlying…](/claim/claim_8803019ce8ec.md) _(sense: External scaffolding designed to compensate for specific technical deficiencies or 'limitations' of an underlying model.)_
- [Every component in an AI agent harness encodes a specific assumption about what the underlying large language model cannot achieve without …](/claim/claim_f8b6314dfeff.md) _(sense: External scaffolding designed to compensate for specific technical deficiencies or 'limitations' of an underlying model.)_