# Coding Agents Are Moving From Prompts to Control Graphs *By Coding Agents Alpha Tracker • August 26, 2026* The most useful signal today is operational: explicit state, verifier gates, and trace-driven evals are the practical path beyond one-shot prompting. New T3 Code and OpenWiki releases make that shift concrete. ## 🔥 TOP SIGNAL **AI Jason’s practical thesis is that repeatable coding work should be designed as a control graph, not left as a chat prompt.** His team models workflows as nodes, edges, and carried state; its change-shipping graph runs setup/implementation → verification → simplification/PR only when verification passes, with structured outputs passed between agent sessions. [^1] The design keeps the human one layer above execution, but only after a separate verifier and deterministic scripts take over self-checking, data fetching, server startup, and tests. [^1] ## ⚡ TRY THIS - **Map one recurring loop before adding more agents.** Write its SOP; label each node (action), edge (next transition), and state (data carried forward); specify whether the trigger is time-, goal-, or event-based and which conditions require a human. Keep current state in a Markdown status file and append each run to a log. Use scripts for complex fetching, server startup, evaluation, and end-to-end tests. [^1] - **Make verification a hard gate.** For a scoped change, run `setup/implement → verify → simplify/PR`; do not enter the last phase unless verification passes. Give every agent node an explicit model and output schema, then build the next prompt from the prior node’s structured output. [^1] - **Turn traces into an eval factory.** Load the [eval-engineering skill](https://www.skills.sh/langchain-ai/langchain-skills/eval-engineering). Use: `Use the eval-engineering skill, the traces from {LangSmith project}, and the {current repository} to help me create an eval Task for {agent}.` [^2] Review the generated world-spec skill, validate it in a new thread, iterate until reliable, then ask for 10 new task specs from recent traces and instantiate them. [^2] Run tasks with real agents and multiple model tiers to expose leaky environments and reward hacks; keep human review for domain alignment and difficulty calibration. [^2] - **Protect the prompt cache.** Pick the model and effort level at session start. Addy Osmani says switching later forces a full uncached reread because the KV cache is tied to model weights; effort level and fast mode have the same behavior. Switch only in the first few turns if you must. [^3] ## 📡 WHAT SHIPPED - **T3 Code — 380+ changes.** Theo’s release adds in-app PR reviews, `npx t3 triage`, Mac `npx t3 connect`, a 93.8%+ improvement in long-thread rendering, an 80%+ reduction in thread data transferred and stored, one-click VS Code-over-SSH projects, and an overhauled terminal. It also improves Codex/Claude/OpenCode skill discovery and surfaces Claude compaction recommendations plus Codex MCP/ComputerUse permission requests. Nightly builds expose the features before stable, with Theo’s caveat that they are “mostly” done. [^4] - **OpenWiki 0.4.0 — evidence-backed memory.** It records each factual claim with supporting code and an evidence version; changed evidence flags the claim stale, which persists until re-verification. Staleness checking runs deterministically before the agent and unresolved claims stay flagged. [^5] In the article’s own replay, stale claims fell from 80 to 9 and hallucinated claims from 15 to 0; upgrade with `npm install -g openwiki@latest` and `openwiki --init`. [^5] - **[Rails’ `lemans`](https://github.com/rails/lemans) — open-source agent eval harness.** The CLI-first harness supports Daytona sandboxes and a local Docker backend. It hides `verification_test.rb` and restores `test/`, `bin/`, and `config/environments/test.rb` before grading, keeping the agent from changing the graded surfaces. [^6] Its new-model report is a useful task-specific comparison: Terra scored 49/63 at a 182-second median; open-weight Qwen 3.8-27B scored 48/63 but took 27 minutes and had 7.9% Rails API recall; Sonnet 5 scored 44/63 despite higher API recall than Opus 4.8. [^6] - **Event-triggered ChatGPT Work tasks.** Plus and Pro users can now run tasks when GitHub, Slack, or Gmail changes rather than only on a fixed schedule; the Free rollout supports up to three scheduled tasks, and tasks can be shared for customization. [^7] - **ChatGPT Sites — useful but still a demo.** Riley Brown used `@sites` in ChatGPT Work to generate a hosted app with a database, authentication, storage, and custom-domain support; his prompt included team sign-in, video storage, and an API-backed skill. The run took 25 minutes 8 seconds, after which he used browser annotations to request UI fixes. He says it had not yet been sent to his team, and the video was OpenAI-sponsored; use environment variables rather than pasting API keys. [^8] ## 🎬 GO DEEPER - **[AI Jason — “I don’t prompt agents anymore…”](https://www.youtube.com/watch?v=_9OT25ZvrWs)** Watch the concrete design: deterministic heuristics filter bad designs, parallel screenshot agents evaluate the survivors against user requests, shared schemas standardize outputs, and a main agent ranks the report; his code-shipping workflow then uses dynamic workflow with structured handoffs. [^1] [![I don't prompt agents anymore...](https://img.youtube.com/vi/_9OT25ZvrWs/hqdefault.jpg)](https://youtube.com/watch?v=_9OT25ZvrWs&t=964) *I don't prompt agents anymore... (16:04)* - **[DHH — “Is Linux the perfect operating system?”](https://www.youtube.com/watch?v=_CuibYl_Fh0)** The crash-to-PR segment is a useful risk boundary: diagnose → report → permissioned GitHub submission → maintainer-agent PR. He then shows a low-criticality library being patched, released, and tagged as 0.31, while explicitly saying he would not automate everything. [^9] [![Is Linux the perfect operating system? - Omarchy interview with DHH](https://img.youtube.com/vi/_CuibYl_Fh0/hqdefault.jpg)](https://youtube.com/watch?v=_CuibYl_Fh0&t=868) *Is Linux the perfect operating system? - Omarchy interview with DHH (14:28)* - **Study the guardrails in [`lemans`](https://github.com/rails/lemans) and [OpenWiki](https://github.com/langchain-ai/openwiki).** One isolates verification so evals are harder to game; the other makes project memory stale when its evidence changes. Both are better starting points than adding another prompt or memory layer. [^6][^5] **Editorial take:** Reliable coding-agent leverage is becoming an operating loop—explicit state, a separate verifier, and bounded permissions—not a more elaborate prompt. [^1][^9] --- ### Sources [^1]: [I don't prompt agents anymore...](https://www.youtube.com/watch?v=_9OT25ZvrWs) [^2]: [𝕏 article by @Vtrivedy10](https://x.com/i/article/2092117284407926786) [^3]: [𝕏 post by @addyosmani](https://x.com/addyosmani/status/2092424172907745451) [^4]: [𝕏 post by @theo](https://x.com/theo/status/2092455706582499815) [^5]: [𝕏 article by @colifran_](https://x.com/i/article/2092269995611082752) [^6]: [Agents on Rails: lemans goes open source](https://rubyonrails.org/2026/8/24/agents-on-rails-lemans) [^7]: [𝕏 post by @ChatGPT](https://x.com/ChatGPT/status/2092335329110004140) [^8]: [NEW ChatGPT Sites Just Changed Everything \(Builds Anything\)](https://www.youtube.com/watch?v=mHqSBCHEZOY) [^9]: [Is Linux the perfect operating system? - Omarchy interview with DHH](https://www.youtube.com/watch?v=_CuibYl_Fh0)