# Disposable Agent Code, Local DS4, and Open Models Passing the Swap Test *By Coding Agents Alpha Tracker • May 9, 2026* The practical signal today is boundary-setting: use agents aggressively on cheap-to-regenerate code, keep context surfaces tight, and pay attention as local and open-model stacks become viable for serious coding workflows. ## 🔥 TOP SIGNAL - **The highest-signal shift today: treat agent-written code as disposable scaffolding when change is cheap.** Mitchell Hashimoto says AI "slop" is useful because it enables fast parallel experimentation; he used agent loops in Ralph overnight to generate dozens of low-quality plugins so he could test a full GUI and plugin ecosystem, then regenerate them whenever the API changed because the cost of change was just tokens [^1]. - **Kent C. Dodds is using the same boundary in practice.** His MCP-powered assistant Kody has produced 160k+ lines of code he has not read, which is acceptable for proving the idea works—but not for a finished product, where he says he would rewrite more intentionally from scratch [^2]. Simon Willison's version is narrower but aligned: he already trusts Claude Code for routine production tasks, while warning that repeated success can create "normalization of deviance" and that real usage matters more than AI-generated tests/docs alone [^3]. ## ⚡ TRY THIS - **Create a prototype-only lane for unstable surfaces.** Mitchell's pattern is concrete: keep core internals high-quality, but let agents generate the GUI, plugins, or provider layer while the API/SDK is still moving; run loops overnight; regenerate on API changes; only ship that slop transparently to testers; then rewrite once the concept is proven [^1][^2]. - **Ask for HTML, not Markdown, when you need an explanation you can inspect.** Thariq Shihipar's argument: HTML lets Claude produce SVG diagrams, interactive widgets, and in-page navigation for code explanations and reviews [^4]. Simon Willison highlighted this exact PR-review prompt: ```text Help me review this PR by creating an HTML artifact that describes it. I'm not very familiar with the streaming/backpressure logic so focus on that. Render the actual diff with inline margin annotations, color-code findings by severity and whatever else might be needed to convey the concept well. ``` [^4] Start with that pattern for gnarly diffs, then browse examples at [thariqs.github.io/html-effectiveness](https://thariqs.github.io/html-effectiveness/) [^4]. - **Stop auto-loading every skill/tool into context.** Dillon Mulroy says he almost never wants skills auto-invoked, so he built custom tooling to toggle them on and off to keep them out of the context window unless needed; Armin Ronacher's terse alternative: prompt templates [^5][^6]. Practical takeaway: keep reusable workflows off by default, enable them only for the current task, and use templates for repeat jobs. - **For code search, start simple: agentic retrieval first, parallel fan-out second, embeddings later.** swyx says simple agentic RAG is good enough for many codebases—especially homogeneous ones—until you're dealing with something on the order of 10B-1T tokens; for search itself, fan out in parallel, e.g. four rounds of eight searches, instead of one-search-at-a-time contexting [^7]. Then add semantic indexing for larger codebases, where tools like Cursor's embedding flow start to matter more [^7]. ## 📡 WHAT SHIPPED - **DS4 for DeepSeek Flash V4** — Salvatore Sanfilippo released DS4, an open-source inference engine for DeepSeek Flash V4 and his first major OSS project built primarily with AI-written code under human architectural control [^8]. He says DeepSeek Flash V4 gives him a usable **1M-token** local context window on **198GB RAM** [^8]. Benchmarks he cited: **~470 t/s prefill + 35-36 t/s generation on M3 Ultra**, **~250/25 on M3 Max** [^8]. - **DS4's practical feature set looks agent-ready** — HTTP API server, streaming, tool streaming via PR, logging, tracing, and a disk-persisted KV cache treated as a first-class object [^8]. The setup flow is unusually short: clone the repo, run `make`, download the model, start the server, then use it with pyagent / cloud code / open code; experienced pyagent collaborators told him that workflow felt like "product mode" rather than a toy [^8]. - **Open-model swap test passed in production** — Caspar Brun says his org changed Fleet's internal model from **Sonnet 4.6** to **Kimi K2.6** and he "didn't even notice"; his claim is that open models are already good enough for most tasks outside the hardest coding work, at **5-10x lower cost** [^9]. LangChain's framing: this is the year of open-source LLMs in agents [^10]. - **Fleet added per-agent tracing control** — You can now enable or disable tracing at the individual agent level in Fleet, which Brace Sproul called a "big unlock" for getting full trace details only where you need them. Docs: [langchain.com/langsmith/fleet](http://langchain.com/langsmith/fleet) [^11][^12]. - **Codex migration got a direct path** — ChatGPT now exposes a [switch-to-Codex flow](https://chatgpt.com/codex/switch-to-codex/); Tibo's practitioner summary was simple: "You can just migrate things" [^13][^14]. - **Current model chatter from an agent lab** — swyx's coding-model shortlist right now: **Claude 4.6** and **GPT-5.3 Codex** [^7]. ## 🎬 GO DEEPER - **10:18-12:34 — swyx on when simple agentic RAG is enough.** Good calibration if you're overbuilding retrieval: for many codebases, plain agentic search works fine until the corpus gets truly large or heterogeneous [^7]. [![AI & Software Quality with Shawn Wang (aka swyx)](https://img.youtube.com/vi/W5SIRbp0AnU/hqdefault.jpg)](https://youtube.com/watch?v=W5SIRbp0AnU&t=618) *AI & Software Quality with Shawn Wang (aka swyx) (10:18)* - **15:44-16:45 — swyx on parallel search.** Watch this if your agent still does naive sequential retrieval; the useful bit is the fan-out pattern—multiple searches per round plus diversity across them [^7]. [![AI & Software Quality with Shawn Wang (aka swyx)](https://img.youtube.com/vi/W5SIRbp0AnU/hqdefault.jpg)](https://youtube.com/watch?v=W5SIRbp0AnU&t=943) *AI & Software Quality with Shawn Wang (aka swyx) (15:43)* - **10:02-11:34 — antirez on the DS4 setup loop.** The benchmark numbers are nice, but the real hook is the workflow: clone, `make`, download model, run server, connect your coding agent stack [^8]. [![DS4: ora su GitHub](https://img.youtube.com/vi/9AjeHE_VReY/hqdefault.jpg)](https://youtube.com/watch?v=9AjeHE_VReY&t=602) *DS4: ora su GitHub (10:02)* - **Study [simonw/tools](https://github.com/simonw/tools).** It contains the kind of narrow, useful artifacts that Claude Code is good at building quickly, including the Redis Array Playground and other small utilities [^3]. - **Study [inaturalist-clumper](https://github.com/simonw/inaturalist-clumper) + [simonw/inaturalist-clumps](https://github.com/simonw/inaturalist-clumps).** This is a clean end-to-end pattern worth copying: small Python CLI → git-scraped JSON → HTML frontend generated from a precise prompt against real data [^3]. *Editorial take: today's edge is boundary-setting—disposable prototype code, tighter context windows, and better search strategy beat blindly giving agents more rope* [^1][^5][^7]. --- ### Sources [^1]: [𝕏 post by @mitchellh](https://x.com/mitchellh/status/2052397933522506079) [^2]: [𝕏 post by @kentcdodds](https://x.com/kentcdodds/status/2052764893586305491) [^3]: [Vibe coding and agentic engineering are getting closer than I’d like](https://simonw.substack.com/p/vibe-coding-and-agentic-engineering) [^4]: [Using Claude Code: The Unreasonable Effectiveness of HTML](https://simonwillison.net/2026/May/8/unreasonable-effectiveness-of-html) [^5]: [𝕏 post by @dillon_mulroy](https://x.com/dillon_mulroy/status/2052778906516619569) [^6]: [𝕏 post by @mitsuhiko](https://x.com/mitsuhiko/status/2052782530525348243) [^7]: [AI & Software Quality with Shawn Wang \(aka swyx\)](https://www.youtube.com/watch?v=W5SIRbp0AnU) [^8]: [DS4: ora su GitHub](https://www.youtube.com/watch?v=9AjeHE_VReY) [^9]: [𝕏 post by @caspar_br](https://x.com/caspar_br/status/2052817936344400132) [^10]: [𝕏 post by @LangChain](https://x.com/LangChain/status/2052819061436973231) [^11]: [𝕏 post by @BraceSproul](https://x.com/BraceSproul/status/2052785315186897316) [^12]: [𝕏 post by @LangChain](https://x.com/LangChain/status/2052785736106299521) [^13]: [𝕏 post by @OpenAI](https://x.com/OpenAI/status/2052800507727781979) [^14]: [𝕏 post by @thsottiaux](https://x.com/thsottiaux/status/2052808587429523663)