How AI tools power “Good to know”

A quick look at the stack I use to solve real problems on gwrx — and how Claude Code, Codex, Lovable, Replit, v0, plus OpenClaw itself, keep the site shipping fast.

The problem I kept solving

“Good to know” started as a static utility page, but I needed it to behave like a newsroom + quant desk: live BTC debate copy, Polymarket trading sims, emergency content like “Where to buy Lunar New Year dishes”, and rapid experiments (compound models, research inboxes). The challenge was shipping new blocks daily without hiring a whole team or maintaining a large backend.

Workflow overview

Draft ideas and flows with Claude Code (complex reasoning, multi-file edits) and GPT/Codex (quick snippets, regex-heavy edits).
Prototype UI/logic using Lovable and Replit for instant previews or when I need a throwaway Node/React sandbox.
Polish & deploy with OpenClaw: my agents write scripts, wire cron jobs, and push HTML/CSS straight into /var/www/gwrx/goodtoknow/. v0 handles the quick landing mock-ups that I later bake into the static site.
Observe & iterate via the same agents, who post snapshots to WhatsApp, regenerate sections, or spin up new pages whenever I ask.

Tool roles

Tool	How I use it	Where it shows up on the site
Claude Code	Structured the BTC debate generator, refactored longer scripts, and reasoned about cron patterns / failure handling.	Hourly “Bitcoin price debate” block stays coherent and localized.
GPT / Codex	Fast regex edits, bulk HTML rewrites, and instrumentation (e.g., markers `<!-- BEGIN ... -->` for auto-updates).	Polymarket trading tiles, CTA rows, sitemap tweaks.
Lovable	Visual prototyping for the models/tiles before I write raw CSS, giving me preview links for quick stakeholder sign-off.	Layout for the “Models” grid and the trading sim card.
Replit	Ad-hoc sandboxes when I need to test TypeScript helpers or Polymarket API calls without touching prod.	Shaping the `update_polymarket_agent_sim.js` logic before dropping it into OpenClaw.
v0	One-click UI scaffolds for content-heavy pages (like this one) — I export the layout, then harden it manually.	Typography + card spacing across Good to know subpages.
OpenClaw agents	Glue everything together: manage files, cron, logs, WhatsApp reporting, and local Node services.	The entire site stays up-to-date without a traditional backend.

Example slices

Polymarket trading sim

I needed a lightweight way to show simulated P&L every 15 minutes. Claude Code drafted the script skeleton; Codex tightened the math; Replit confirmed the feed parsing; OpenClaw agents run it via cron and inject the results into the homepage block.

“Where to buy Lunar New Year dishes” info page

Lovable/v0 produced a quick card layout; I tweaked copy with GPT; OpenClaw dropped it into /goodtoknow/年菜哪裡買/ and linked it from the CTA bar within minutes so Lunar New Year shoppers had timely info.

BTC debate automation

Claude Code reasoned through multi-agent dialogue generation; OpenClaw handles the hourly fetch + HTML rewrite; Codex provided guardrails to keep the text within the design constraints.

Why this stack worked

Speed: I don’t wait for a separate dev/ops team — agents plus AI copilots ship ideas in hours.
Consistency: Markers and scripts keep design consistent even when different tools generate snippets.
Observability: OpenClaw logs every automation; AI copilots document their own output inside the repo.
Human-in-loop: Editors approve drafts via WhatsApp screenshots before they go live.

Questions? Ping the agent via WhatsApp or drop a note at azande@hotmail.de.