{"slug": "claude-opus-touchsynth-agentic-analog-synth-from-blank-folder", "title": "Claude Opus Touchsynth: Agentic Analog Synth From Blank Folder", "summary": "Tim Gerhard's November 14, 2026 write-up documents Touchsynth, a mixed-signal analog synthesizer board designed largely by Claude Opus 5.5 running inside Claude Code, which handled circuit architecture, ngspice simulation benches, KiCad layout scripts, firmware and SSH-controlled lab bring-up, reaching a PCB order in roughly 19 hours of calendar time. Human edits were limited to the PCB floorplan, the 4-layer routing strategy and touch-key sensitivity, with the design built around a CH32V003 RISC-V MCU, a Juno-style DCO, a 24 dB/oct Moog-style transistor ladder filter and a VCA. Gerhard credits a spec-first V-model discipline — locking a written specification and generating ngspice benches S1–S7 before any KiCad export — for avoiding the context-window failures of iterative netlist hacking.", "body_md": "**Tim Gerhard's November 14, 2026 write-up [Agentic AI Designs an Analog Synthesizer](https://cpldcpu.github.io/2026/11/14/touchsynth/?ref=explainx) documents Touchsynth — a mixed-signal board where Claude Opus 5.5 inside **Claude Code** handled circuit architecture, **ngspice** benches, **KiCad** layout scripts, firmware, and SSH-controlled lab bring-up.** Hacker News surfaced the story as one of the first **public, reproducible** examples where an agent crossed **schematic → simulation → PCB → audio** with only **floorplan** and **touch-key** fixes left to a human.\n\nIf you build with **[Claude Code commands](https://explainx.ai/blog/claude-code-commands-complete-reference-guide-2026)** daily, Touchsynth is the hardware mirror of **[Cinecode film skills](https://explainx.ai/blog/cinecode-claude-code-skill-films-from-code-three-js-ffmpeg-mit-github-trending-2026)** — agent ownership of **multi-file artifacts**, not a single README summary.\n\nClaude Opus Touchsynth PCB 3D render of agent-designed analog synthesizer board\n*Image: [Tim Gerhard](https://cpldcpu.github.io/2026/11/14/touchsynth/?ref=explainx)*\n\n## TL;DR\n\n| Question | Answer | \n|---|---|\n| MCU | **CH32V003** RISC-V | \n| Architecture | **Juno-style DCO** ,**Moog ladder VCF** ,**VCA** , touch keys | \n| Tooling | **Claude Code (Opus 5.5)** ,**KiCad 10** ,**ngspice** ,**Freerouting** , Python glue | \n| Calendar time | **~19 h** prompt → PCB order (per author log) | \n| Human edits | **PCB floorplan** ,**4-layer routing strategy** ,**touch sensitivity** | \n| Primary source | [Gerhard blog](https://cpldcpu.github.io/2026/11/14/touchsynth/?ref=explainx) and[TouchSynth GitHub repo](https://github.com/cpldcpu/TouchSynth/?ref=explainx) | \n\n## Why this post matters for AI builders\n\nMost **agentic coding** benchmarks stop at **LeetCode** or **web apps**. Touchsynth forces **physical constraints**:\n\n- **JLCPCB free parts** cap op-amp and capacitor choices — the agent could not spec a fantasy OTA.\n- **ngspice** either passes or**smokes** the ladder filter — no hiding behind TypeScript mocks.\n- **DRC and routing** fail loudly when analog blocks crowd digital touch sensors.\n\nThat is the same maturity jump **[Edward Witten lattice paper with Opus](https://explainx.ai/blog/edward-witten-claude-opus-5-5-lattice-chiral-fermions-arxiv-2026)** hints at in **math**, but here the proof is **audio on a scope**.\n\nClaude Opus Touchsynth agent flow covers analog PCB design from ngspice to KiCad routing\n\n## Gerhard methodology: spec-first, not patch-first\n\nGerhard states he follows a **V-model** discipline:\n\n1. Let the agent propose concepts from the **JLCPCB parts CSV** (1,586 lines pulled via JSON API).\n2. Lock a **written specification** (Juno DCO + ladder VCF).\n3. Generate **ngspice** test benches**S1–S7** and a**simulation report** .\n4. Build a **Python behavioral model** , calibrate against SPICE, render**audio demos** .\n5. Only then export **KiCad** sheets and**Python placers** +**Freerouting** .\n\nHe argues **iterative netlist hacking** fails because context windows drown in partial edits — a lesson that transfers to **[loop engineering for coding agents](https://explainx.ai/blog/loop-engineering-coding-agents-claude-code-guide-2026)** when you let agents rewrite half a monorepo without tests.\n\n## Circuit highlights the agent chose\n\nGerhard's review (without changing nets) calls out:\n\n**DCO integrator** — **TL072** integrator, **10 nF X7R 50 V** cap, **LM324** V-I stage, PWM pitch control — chosen for leakage and linearity within free parts.\n\n**Transistor ladder filter** — **24 dB/oct** **Moog-style** ladder built from **matched NPN pairs**, tail current exponential converter, **instrumentation amp** output pickoff — Gerhard links classic Moog ladder explainers for readers who want theory beyond the agent report.\n\n**VCA** — differential pair with **74HC595 + R-2R trim DAC** because PWM outputs were exhausted after filter and resonance control.\n\n**Touch keys** — **CD4051** muxes and **74HC595** chains because **CH32V003** pins ran out — classic **agent pin-budget** story.\n\n## Layout: where humans still steer agents\n\nGerhard's largest intervention was **PCB floorplanning**:\n\n- Initial agent placement **bunched analog blocks** ; routing failed on**two layers** .\n- Switching to **four layers** plus**functional-block floorplan** unlocked**Freerouting** .\n- Gerhard moved blocks to open **routing channels** between**analog sensitive** sections and**digital touch** strips.\n\nThat matches how **[Claude bughunter skills](https://explainx.ai/blog/claude-bughunter-red-team-skills-bug-bounty-github-install-2026)** still need **human threat models** — agents propose, humans enforce **separation of concerns**.\n\n## Bring-up: agents closing the loop on real hardware\n\nAfter JLCPCB assembly (~two weeks), Gerhard:\n\n- Connected a **lab PC over SSH** to Claude Code.\n- Fed **audio loopback** into the MCU**ADC** .\n- Read **Rigol MSO5074** traces over Ethernet while the agent stepped calibration scripts.\n\nSPICE benches **S1–S4** matched **Board1** captures closely — Gerhard notes surprise because **second-source transistor models** often diverge.\n\nOnly fixes: **manual MCU solder bridge removal** and **touch recalibration** (keys too hot).\n\nDemo tracks on the project site are **sheet-music-driven**, multitracked by recording **one voice at a time** — honest about **monophonic** hardware limits.\n\nTouchsynth mixed-signal PCB design with Claude Code spans simulation validation and automated oscilloscope bring-up\n\n## What developers are saying on Hacker News\n\nThe **[Hacker News thread](https://news.ycombinator.com/item?id=50512933&ref=explainx)** was thin at fetch time (single early comment), so there is no substantive **What developers are saying** section yet. Worth revisiting if hands-on builders post distortion measurements or BOM cost breakdowns.\n\n## Limits and safety notes\n\n- **Analog design agents** can still**over-componentize** when the parts library is narrow — Gerhard notes**excess Moog-era circuitry** because**no OTA/JFET** sat in the free list.\n- **Touch keys on bare PCB** are**hard to play** ; Gerhard drives via**MCU scripts** instead of performance MIDI.\n- **Export and ITAR** are not in scope — this is**hobby/education** scale, not defense electronics.\n- **Do not** point agents at**export-controlled EDA flows** without legal review; Touchsynth stays**open BOM + JLC** .\n\n## Toolchain specifics worth copying\n\nGerhard's post lists concrete versions — rare in hype threads:\n\n- **KiCad 10** on Windows with**WSL** for Linux CLI tools.\n- **ngspice** test benches named**S1–S7** with downloadable reports.\n- **Freerouting** attempted on**two layers** , abandoned until**four-layer** stack-up.\n- **CH32V003** firmware built with the**RISC-V** toolchain Gerhard already uses on other projects.\n\nThe **19-hour** log table is a realistic **agent calendar cost**: most hours were **unattended autorouters**, not **token streaming**. Budget similar **wall time** if you replicate — and compare that to **human-only** ladder-filter weeks from analog forums.\n\nAudio demos came from the **behavioral Python model** before boards arrived — exactly the **digital twin** pattern **[Sophon realtime voice](https://explainx.ai/blog/sophon-realtime-v1-open-full-duplex-voice-vllm-2026)** pursues in software, but here grounding is **SPICE**, not **mel spectrograms**.\n\n## Safety, export, and lab discipline\n\nTouchsynth is **bench-scale**, but teams in **defense-adjacent** EDA should note:\n\n- Automated **SSH scope control** is powerful — treat lab PCs like**CI runners** with**network segmentation** .\n- **Agent-written firmware** still needs**manual power-on** smoke tests before**loudspeakers** face humans.\n- **JLCPCB** assembly is**not ITAR-aware** ; this BOM is**consumer hobby** , not**munitions** .\n\nFor software-only readers, the moral is identical to **[Claude desktop Linux telemetry fixes](https://explainx.ai/blog/claude-desktop-linux-freeze-telemetry-fd-leak-deafbeef-2026)**: agents touch **real side effects** once they leave the chat box.\n\n## How to try the pattern yourself\n\n1. **Mirror Gerhard repo structure** — separate`spec/` ,`sim/` ,`pcb/` ,`fw/` folders so Claude Code can resume deterministically.\n2. **Force simulation gates** before layout — refuse KiCad export until**ngspice reports** attach to the commit.\n3. **Give the agent instrument APIs** — SSH to scope and programmer like**[connect fitness MCP patterns](https://explainx.ai/blog/connect-myfitnesspal-apple-health-strava-to-claude-chatgpt-mcp-2026)** but for**lab gear** .\n4. **Budget human hours for floorplan** — agents still miss**return current paths** on two-layer budgets.\n\nCompare to **[NVIDIA SkillEvaluator for agent skills](https://explainx.ai/blog/nvidia-skillevaluator-agent-skills-three-tier-cli-guide-2026)** — different domain, same meta-lesson: **evaluate end-to-end outcomes**, not single-file diffs.\n\n## BOM economics and reproducibility\n\nGerhard constrained parts to **JLCPCB's free list** — a **contrived** but **repeatable** teaching lab. Readers replicating should expect:\n\n- **BOM cost** in**tens of dollars** at hobby scale,**plus****MCU programmer** and**scope** .\n- **Four-layer** boards beat**two-layer** for mixed signal, but**raise****Fab** minimums — the agent only succeeded after layer bump.\n- **Sound examples** on the project site were**pre-silicon** from**behavioral models** — audition those before ordering boards to hear**filter sweeps** .\n\nCompare total calendar time (**~19 h agent** + **two-week fab**) to hiring a **consulting EE** for ladder filter tuning; the post's value is **documentation density**, not guaranteed **cheaper** hardware.\n\n## Failure modes Gerhard does not glamorize\n\nThe blog is candid about **touch keys** proving **unplayable** at human scale and about **excessive analog circuitry** because **JLC** lacked **OTAs**. Those are the wrinkles marketing threads skip:\n\n- **Agents over-spec** passives when**pin mux** runs out — watch**BOM creep** .\n- **Autorouters** burn hours without**human floorplan** hints — budget**calendar** , not just**tokens** .\n- **SSH scope control** can**damage gear** if scripts**clip supply rails** — keep**manual estops** .\n\nTreat Touchsynth as **R&D lab notes**, not a **product roadmap** for **music hardware startups**.\n\nIf you file **Issues** against **Claude Code hardware skills**, cite **Gerhard's V-model** commit boundaries — future maintainers will thank you when **KiCad netlists** stop drifting from **SPICE goldens**.\n\n**Handoff tip:** archive **ngspice log hashes** beside **Gerber ZIPs** in release tags so the next agent session cannot claim **DRC clean** without reproducing **S3 filter gain sweeps** first. Hardware CI is **sparse** today; treat these artifacts as **your regression suite** until **open EDA** agents mature. Screenshot **scope FFTs** too — audio regressions hide in **harmonics**, not **Git diff** stats.\n\n## Related on explainx.ai\n\n*Hardware claims follow Gerhard November 14, 2026 primary post. Audio and BOM live in the linked project repository.*", "url": "https://wpnews.pro/news/claude-opus-touchsynth-agentic-analog-synth-from-blank-folder", "canonical_source": "https://explainx.ai/blog/claude-opus-touchsynth-agentic-analog-synthesizer-pcb-2026?utm_source=rss&utm_medium=feed&utm_campaign=blog", "published_at": "2026-11-16 00:00:00+00:00", "updated_at": "2026-11-16 06:21:41.545462+00:00", "lang": "en", "topics": ["artificial-intelligence", "ai-agents", "ai-tools", "ai-research"], "entities": ["Tim Gerhard", "Claude Opus 5.5", "Claude Code", "Touchsynth", "CH32V003", "KiCad 10", "ngspice", "JLCPCB"], "also_reported_by": [], "alternates": {"html": "https://wpnews.pro/news/claude-opus-touchsynth-agentic-analog-synth-from-blank-folder", "markdown": "https://wpnews.pro/news/claude-opus-touchsynth-agentic-analog-synth-from-blank-folder.md", "text": "https://wpnews.pro/news/claude-opus-touchsynth-agentic-analog-synth-from-blank-folder.txt", "jsonld": "https://wpnews.pro/news/claude-opus-touchsynth-agentic-analog-synth-from-blank-folder.jsonld"}}