Using Electrum without Claude Code — just the templates by hand?

[yoelf22]

My company doesn't allow AI tools on proprietary product designs. But the templates and checklist look useful on their own. Has anyone run the process manually? How long does it take, and which phases are most valuable without AI assistance?

Also: is there a recommended order if I only have time for 2-3 phases instead of all 8?

[yoelf22]

This is a valid way to use the toolkit — the templates predate the Claude Code skill.

Time estimates for manual use

Phase	With AI	Manual	Notes
1. Explore	30 min	2–3 hours	The skills map (16 areas) takes time to work through honestly
2. High-Level Design	20 min	4–6 hours	Drawing the block diagram by hand is actually useful — forces precision
3. Component Arrangement	20 min	3–4 hours	Sketching spatial layout on paper or in CAD
4. System Description	30 min	1–2 days	This is where manual effort is highest — naming real parts, computing power budgets
5. Gate Checklist	15 min	2–3 hours	90 items, each needs a yes/no/skip decision
6–8. Visuals + Carousel	20 min	Optional	Skip these if you're not presenting externally

Total manual: roughly 3–5 working days spread over 1–2 weeks. The AI compresses this to ~90 minutes, but the manual version often produces better results because you're forced to sit with each decision longer.

If you only have time for 2–3 phases

Prioritize in this order:

1. System Description (Phase 4) — this is the highest-value single artifact. Copy templates/system_description_template.md into your project folder and fill in every section. The act of writing forces you to confront gaps. When you hit a section you can't fill, you've found a risk.

2. Gate Checklist (Phase 5) — run through templates/checklist.md after the system description. The checklist is specifically designed to catch what the system description misses. Pay special attention to items tagged [HW↔SW] — those are the cross-domain gaps that cause EVT surprises.

3. Explore (Phase 1) — if you have time for a third, start here. The skills map (templates/skills_map.md) is a 30-minute exercise that tells you which competency areas your team covers and which are gaps. This is strategic, not technical.

What you lose without AI

The main thing AI adds is adversarial specificity. When you write a system description yourself, you naturally gloss over areas you're less confident about. The AI pushes back: "You said BLE connectivity but didn't specify the connection interval, MTU size, or reconnection strategy." You can replicate this by having a colleague review each phase output before you move to the next one — the process is designed for that kind of back-and-forth.

The block diagram and arrangement diagram phases (2 and 3) also benefit from AI because generating matplotlib visualizations by hand is tedious. For manual use, a whiteboard sketch or a simple draw.io diagram works fine — the format doesn't matter, the completeness does.