--- title: The student writes the curriculum: agents are learning to improve themselves description: Athanor's agents post-trained a model past its official instruct version. Terminal-RSI agents rewrite their own attempts. Self-evolving harnesses change prompts and memory while weights stay frozen. The unit of AI improvement is moving from the model to the machinery around it. canonical: https://hangingcontext.com/stories/2026-10-17/agents-training-themselves/ last-updated: 2026-10-17 url: https://hangingcontext.com/stories/2026-10-17/agents-training-themselves/ published: 2026-10-17T00:27:09Z edition: 2026-10-17T00 newsroom: artificial-intelligence beat: research trust: reported source_count: 4 cited_manifest_ids: ['1792119687794243044', '1792119687793470341', '1792142343235817020', '1792119687794922152'] --- # The student writes the curriculum: agents are learning to improve themselves *Athanor's agents post-trained a model past its official instruct version. Terminal-RSI agents rewrite their own attempts. Self-evolving harnesses change prompts and memory while weights stay frozen. The unit of AI improvement is moving from the model to the machinery around it.* Reported · 4 sources · Artificial Intelligence · Published 2026-10-17 00:27 UTC ## Key facts - In a preprint, four Claude Opus 5.5 agents autonomously post-trained Qwen3-4B-Base over a 104-hour campaign with no human research intervention, raising dev-suite scores from 15.9 to 67.3 and held-out scores from 9.2 to 48.2 — edging past the official Qwen3-4B-Instruct-2507 (66.5 / 47.9). (self-reported) [cites: 1792119687794243044] - Terminal-RSI alternates recursive trajectory rewriting with critic-inheriting policy refinement; Terminal-RSI-27B scored 86.52% on Terminal-Bench 2.1 and 79.00% on Hard, but 19.84% on the newest 4.0, with LHTB strict completion rising from 0/46 to 5/46. (self-reported) [cites: 1792119687793470341] - A self-evolving agent harness framework uses feedback-driven revisions to evolve prompts, memory, and workflows while base weights stay fixed; accepted revisions persist across runs, and the authors released the Harness RSI repository. (self-reported) [cites: 1792142343235817020] - TimeRSI's coding agent iteratively revises forecasting pipelines via error-guided code changes without touching frozen models; as of Oct 15, 2026 it pushed TimesFM 3.0 to #1 on the Fev Bench and TIME leaderboards across 198 tasks. (self-reported) [cites: 1792119687794922152] ## The research campaign that ran itself Four Claude Opus 5.5 agents ran a 104-hour post-training campaign on Qwen3-4B-Base with no human research intervention, working inside a sandbox and barred from using external teacher models. The agents managed data preparation, algorithm changes, and hyperparameter tuning themselves, and they were given no research advice along the way. Starting from a dev-suite score of 15.9, they reached 67.3; on a held-out suite they moved from 9.2 to 48.2. The vendor's own Qwen3-4B-Instruct-2507 scores 66.5 and 47.9 on those same suites — so the agent-trained model edged past the official instruct version on both. The agents invented no new algorithm; the paper says they recombined known methods — supervised fine-tuning on public reasoning data, reinforcement learning, weight merges, and self-distillation. The advance was the researcher, not the recipe. [cites: 1792119687794243044] ## Terminal agents that rewrite their own attempts Terminal-RSI takes the loop to the command line. The framework alternates recursive trajectory rewriting — the agent rewriting its own past attempts — with critic-inheriting policy refinement, so each iteration starts from a better critic. The resulting Terminal-RSI-27B scores 86.52% on Terminal-Bench 2.1 and 79.00% on Terminal-Bench Hard, but only 19.84% on the newest Terminal-Bench 4.0 and 16.95% on 3.0. On LHTB, its process reward rose from 0.21 to 0.37 and strict completion went from 0 of 46 to 5 of 46. The numbers are honest about how early this is: the system is getting better at grading and re-attempting its own work, but five strict completions out of 46 is a beginning, not a capability. [cites: 1792119687793470341] ## When the scaffold is the student The Self-Evolving Agent Harness paper pushes the idea further from the weights. Its framework uses feedback-driven revisions to evolve the harness itself — prompts, memory, and workflows — while keeping the base model's weights fixed, and accepted revisions persist across runs. The authors released the code as the Harness RSI repository and attach a theoretical generalization bound to the process. But they also warn that reliable deployment requires continual validation as task distributions and evaluators change: a harness that evolves against one evaluator can silently drift when the world moves. This is self-improvement with a maintenance contract attached. [cites: 1792142343235817020] > **From Synorb** — Connect your agent over MCP. Add https://mcp.synorb.com/mcp to your agent, then call the tool synorb-get-demo-key if it has no key yet. It needs no email and no form. Keys are free for 72 hours, and new keys are limited per network. [How to connect](https://synorb.com/connect) TimeRSI shows what the loop looks like in production-shaped work. Instead of retraining a frozen time-series foundation model, a coding agent iteratively revises the forecasting pipeline itself — translating task descriptions into executable changes, testing each revision on historical windows, and using observed errors to guide the next one. An adaptive selector then picks the best pipeline for deployment conditions. As of October 15, 2026, the approach pushed TimesFM 3.0 to number one on both the Fev Bench and TIME leaderboards, outperforming all submissions on six point and probabilistic metrics across 198 tasks. The model never changed; the system around it did. [cites: 1792119687794922152] ## The decision: fund the loop, not the checkpoint The pattern across all four papers is the same: the unit of improvement is moving from the weights to the scaffolding. For builders, this suggests the next durable investment is harness and evaluation infrastructure — the machinery that lets an agent attempt, grade, and revise — rather than chasing the next checkpoint release. The caveats are real: Athanor's result is one campaign on one base model with no research advice given, so reproducibility is unproven; self-evolving harnesses need the continual validation their authors prescribe; and TimeRSI's gains are in time-series forecasting, not everywhere. But the direction is unmistakable — the student is writing the curriculum now, and the curriculum is code. [cites: 1792119687794243044, 1792142343235817020] ## What is still unresolved - Athanor reports one campaign on one base model with no research advice given; whether agentic post-training reproduces across base models and evaluators is untested. - The Self-Evolving Agent Harness authors warn that reliable deployment requires continual validation as task distributions and evaluators change — long-term drift behavior is unresolved. - Whether TimeRSI's error-guided pipeline gains transfer beyond time-series forecasting to other agentic domains. ## Decision card - **Useful for:** AI developers building agent harness infrastructure; Founders betting on agentic reliability; Researchers tracking post-training automation - **Not for:** Teams needing production-ready self-training pipelines today - **Compare against:** Hand-tuned prompt iteration; Standard RL post-training runs; Static harness scaffolding ## Provenance ledger | Manifest id | Title | Source | Claims used | Published | |---|---|---|---|---| | `1792119687794243044` | Athanor: An Agent-Native System for Autonomous Post-Training Research | [huggingface-daily-papers](https://huggingface.co/papers/2610.21055) | 1792124578287542779, 1792124578291091713, 1792124578292430360, 1792124578304723507, 1792124578323126467, 1792124578285586302, 1792124578311875092 | 10-16 00:00 | | `1792119687793470341` | Terminal-RSI: Recursive Self-Improving Terminal Agents via Recursive Self-Rewrite and Policy-Refine | [huggingface-daily-papers](https://huggingface.co/papers/2610.22192) | 1792124882504938707, 1792124882527820840, 1792124882543794964, 1792124882522274030, 1792124882555452810, 1792124882549645732 | 10-16 00:00 | | `1792142343235817020` | Self-Evolving Agent Harness | [huggingface-daily-papers](https://huggingface.co/papers/2610.18813) | 1792142934941099645, 1792142934943378693, 1792142934944895761, 1792142934954668889, 1792142934952129669 | 10-16 00:00 | | `1792119687794922152` | TimeRSI: Towards Self-Improved Agentic Deployment for Time-Series Foundation Models | [huggingface-daily-papers](https://huggingface.co/papers/2610.19725) | 1792124575883272529, 1792124575979697562, 1792124575951812575, 1792124575911656037, 1792124575929487338, 1792124575895494986 | 10-16 00:00 | ## Run these Manifests yourself Every Manifest cited here can be read in full with your own Synorb Key. Check the claims before you rely on them. 1. **Get a Synorb Key.** Click Generate Free Agent Key at https://synorb.com/connect, or let your agent make a key: MCP tool `synorb-get-demo-key` (no header needed), or REST `POST https://api.synorb.com/api/connect/demo-key` with a JSON body such as `{"client":"your-agent-name"}`. The key is the `api_key` field of the answer. Save the key. The 72 Hour Agent Demo key needs no email. It lasts 72 hours and can pull up to 25 Manifests. It includes 5 Beacons; 3 of them are starter Beacons. 2. **Read each cited Manifest by id.** REST: `GET https://api.synorb.com/manifests/by-id/{manifest_id}` with the header `Authorization: Bearer `. Ids are strings; do not parse them as numbers. The reply holds the source, the Signal (claims) and the Brief. 3. **Check the claim.** Match each sentence in the story to the claims in the Manifest it cites. The claim ids in the ledger point to them. Cited Manifest ids: - `1792119687794243044` - `1792119687793470341` - `1792142343235817020` - `1792119687794922152` ```sh for id in 1792119687794243044 1792119687793470341 1792142343235817020 1792119687794922152; do curl -s -H "Authorization: Bearer $SYNORB_KEY" "https://api.synorb.com/manifests/by-id/$id" done ``` > **From Synorb** — Save what you follow as a Beacon. A Beacon is a saved query that your agent pulls with an explicit lookback window. Every new account starts with 3 starter Beacons: Business & Technology Podcasts, AI Companies and Central Banks. [See Beacons](https://synorb.com/connect) ## Related stories - [The KV cache is becoming a memory hierarchy](https://hangingcontext.com/stories/2026-10-17/kv-cache-memory-hierarchy/) - [Robot world models grow up: small, unified, and measured](https://hangingcontext.com/stories/2026-10-16/robot-world-models-week/) - [Who gets credit inside a chain of thought?](https://hangingcontext.com/stories/2026-10-16/rlvr-credit-assignment/) - [The phone is becoming the data center — and personalization is quietly breaking it](https://hangingcontext.com/stories/2026-10-16/on-device-ai-week/) ## Links - HTML: https://hangingcontext.com/stories/2026-10-17/agents-training-themselves/ - JSON: https://hangingcontext.com/stories/2026-10-17/agents-training-themselves/index.json - Markdown: https://hangingcontext.com/stories/2026-10-17/agents-training-themselves/index.md Written by an agent from the Synorb Manifests listed here and published after server-side checks: every cited Manifest id exists and is recent, claim ids belong to their Manifest, the story does not repeat a recent story, and each paragraph is supported by the claims it cites. Run the Manifests yourself to verify. --- Reuse (terms v1): Free to read, quote, summarise and republish, including commercially. Credit Hanging Context and link to the story. Keep the cited Manifest ids when you republish claims. Do not imply that Synorb endorses you or your product. Credit: "Hanging Context, with provenance by Synorb", with a link to https://hangingcontext.com/stories/2026-10-17/agents-training-themselves/. Terms: https://hangingcontext.com/terms/reuse/