# Muse Code Makes Cost and Locality Part of the Coding-Agent Stack *By Coding Agents Alpha Tracker • August 11, 2026* Meta’s Muse Code beta and Muse Glimmer push cheaper terminal and local execution into the coding-agent stack. Practitioner tests and new orchestration tools point to task-level routing, reusable loops, and reviewable autonomy as the practical edge. ## 🔥 TOP SIGNAL **Meta is making cost and locality first-class coding-agent choices.** Muse Code beta is a terminal agent for complete software-engineering tasks across large repos—planning changes, writing code, and validating results—powered by Muse Spark 1.2; Riley Brown places it between Opus and GPT-5.6 Tera and calculates roughly $5.50 in combined input/output cost for Meta’s top model versus $30 for Opus and $35 for GPT-5.6. [^1] Muse Glimmer adds a 30B Apache 2.0 agentic model: Simon Willison generated an example with LM Studio’s 18.16GB build and says a 32GB+ machine leaves room for other applications. [^2][^3] The practical shift is to measure completed-task cost, latency, and local execution alongside output quality—not simply choose the highest-ranked model. [^4][^5][^1] ## ⚡ TRY THIS - **Route by total task cost, not token price.** In DHH’s follow-on rewrite test, GPT-5.6 Sol High followed Fable’s plan: the first pass was 30% slower, but one follow-up reached parity at $43; DeepSeek V4 Flash could not get anything working despite many follow-ups. [^6] Kimi K3 was stopped after $60; when DHH let K3 Fast finish, it used $80 in tokens—about $55 at standard pricing—and was still slower and more expensive than Sol High. [^7][^8] Start with a fixed acceptance test and a turn/dollar budget for the cheap model, escalate when it stalls, and log total task cost. Riley’s warning is the right accounting rule: low per-token pricing can be erased by extra turns. [^1] - **Make optimization a separate, measured pass.** DHH reports that a GPT Sol optimization run immediately made `ttfx` 63% faster—14× the original `tte`—followed by 16× and then 27× results in later rounds. [^9][^10][^11] After the baseline works, give the agent a bounded prompt such as: `Profile this implementation against the original; optimize only measured bottlenecks; run the same benchmark; report before/after and regressions.` - **Orient before editing a neglected repo.** ThePrimeagen opened a project untouched for more than a year, asked `"yo ai, remind me of ...."`, and got up to speed in about three minutes. [^12] Use that first pass to recover architecture, entry points, tests, and unknowns; do not start implementation until the agent can restate the plan. - **Use a reviewer agent to create a triage map.** Kent C. Dodds asked Devin for a deep review of Kody. The returned review took six minutes and added 96 lines on a roughly 470,000-line codebase, citing zero TODO/FIXME/HACK comments, four uses of `any`, 606 test files, matching security invariants, and a “strong but not perfect” verdict. [^13][^14] Treat this as a fast findings inventory for human verification, not as an autonomous approval gate. ## 📡 WHAT SHIPPED - **Muse Glimmer open weights.** Meta announced a 30B dense model trained for agentic use cases under Apache 2.0, with a stated 24GB-of-VRAM target; the GGUF build is available on Hugging Face. [^2][^4][^15] - **Cross-harness coordination is productizing.** Spotify launched Xirp, a vendor-neutral environment for managing sessions across Claude, Gemini CLI, and Codex; Spotify says 1,300+ engineers already use it and the service is now available to try. [^16] Riley Brown’s Buzz walkthrough shows the complementary team pattern: @mention Codex and Claude Code in one Slack-like thread, create a Cursor agent backed by Kimi K3, add it to channels, and watch which agent is working or requesting approval; the demo agent joined seven channels. [^1] - **Kody v2026.08.10** adds built-in OAuth integrations. Operators can provision shared GitHub, Google, Slack, and similar OAuth apps through `/admin/platform-integrations`; Kent says onboarding feedback made this a necessary shift. [^17][^18] - **`loop-library` is now open source.** Jason Zhou’s collection packages copyable prompts with loops he says deliver real-world results; he says most of the listed loops are already running in SuperDesignDev. [^19][^20] - **Durable agent chat got a concrete product treatment.** Addy Osmani highlights Trigger.dev’s new chat agent, which survives refreshes, crashes, and redeploys and can pause for permission before a risky tool action. [^21] ## 🎬 GO DEEPER - **[Riley Brown — Meta’s NEW Muse Code is Here and Major Codex Updates](https://www.youtube.com/watch?v=piOSL8Ms9mY):** Watch the install → authentication → terminal `muse` → Wii Bowling flow. The important operational detail is the sandbox tradeoff: the demo configures YOLO mode to remove permission prompts and explicitly gives the agent full control, so reproduce that only in a disposable environment. [^1] [![Meta's NEW Muse Code is Here and Major Codex Updates](https://img.youtube.com/vi/piOSL8Ms9mY/hqdefault.jpg)](https://youtube.com/watch?v=piOSL8Ms9mY&t=153) *Meta's NEW Muse Code is Here and Major Codex Updates (2:33)* - **[LangChain — Create an agent that can browse the web with Managed Deep Agents and Browserbase’s Stagehand](https://www.youtube.com/watch?v=O0hkpChFBkM):** Study the tool contract—`snapshot`, `run`, and `screenshot`—including the rule to re-snapshot after navigation; then follow the local `mda dev` test, Browserbase session replay, and `mda deploy` flow. The deployment keeps instructions in Context Hub, so prompts can change without redeploying the agent. [^22] [![Create an agent that can browse the web with Managed Deep Agents and Browserbase's Stagehand](https://img.youtube.com/vi/O0hkpChFBkM/hqdefault.jpg)](https://youtube.com/watch?v=O0hkpChFBkM&t=208) *Create an agent that can browse the web with Managed Deep Agents and Browserbase's Stagehand (3:28)* - **Repo to study — [`ttfx`](https://github.com/omacom-io/ttfx):** Trace the progression from the first measured 63% optimization to 14×, 16×, and 27× claims. It is a compact example of separating agent-generated implementation from repeatable performance passes. [^9][^10][^11] **Editorial take:** The high-alpha move is not picking one permanent model winner; it is making model routing, context recovery, optimization, review, and permissions explicit stages of one inspectable workflow. [^1][^6][^21][^19] --- ### Sources [^1]: [Meta's NEW Muse Code is Here and Major Codex Updates](https://www.youtube.com/watch?v=piOSL8Ms9mY) [^2]: [𝕏 post by @alexandr_wang](https://x.com/alexandr_wang/status/2086756152034066792) [^3]: [Introducing Muse Glimmer](https://simonwillison.net/2026/Aug/10/introducing-muse-glimmer) [^4]: [𝕏 post by @jack_w_rae](https://x.com/jack_w_rae/status/2086798875806015980) [^5]: [𝕏 post by @jack_w_rae](https://x.com/jack_w_rae/status/2086798878096109646) [^6]: [𝕏 post by @dhh](https://x.com/dhh/status/2086783401445482830) [^7]: [𝕏 post by @dhh](https://x.com/dhh/status/2086788943622619443) [^8]: [𝕏 post by @dhh](https://x.com/dhh/status/2086822365485261072) [^9]: [𝕏 post by @dhh](https://x.com/dhh/status/2086774419595083969) [^10]: [𝕏 post by @dhh](https://x.com/dhh/status/2086803426491457976) [^11]: [𝕏 post by @dhh](https://x.com/dhh/status/2086884836120137804) [^12]: [𝕏 post by @ThePrimeagen](https://x.com/ThePrimeagen/status/2086825355013128290) [^13]: [𝕏 post by @kentcdodds](https://x.com/kentcdodds/status/2086893598684152294) [^14]: [𝕏 post by @kentcdodds](https://x.com/kentcdodds/status/2086893921599369466) [^15]: [𝕏 post by @simonw](https://x.com/simonw/status/2086811799480086773) [^16]: [𝕏 post by @SpotifyEng](https://x.com/SpotifyEng/status/2086795659651191106) [^17]: [𝕏 post by @kodykoala](https://x.com/kodykoala/status/2086809070619562298) [^18]: [𝕏 post by @kentcdodds](https://x.com/kentcdodds/status/2086812687108030769) [^19]: [𝕏 post by @jasonzhou1993](https://x.com/jasonzhou1993/status/2086790895538700379) [^20]: [𝕏 post by @jasonzhou1993](https://x.com/jasonzhou1993/status/2086790897489129484) [^21]: [𝕏 post by @addyosmani](https://x.com/addyosmani/status/2086864473139085431) [^22]: [Create an agent that can browse the web with Managed Deep Agents and Browserbase's Stagehand](https://www.youtube.com/watch?v=O0hkpChFBkM)