Rod Bland | Agentic Systems Engineer (@rodbland) header
Rod Bland | Agentic Systems Engineer (@rodbland) profile picture

Ecommerce operator for 20 yrs. Now building AI agent systems for messy ops: automation, dashboards, decision support, and better delivery quality.

Sydney
Joined May 2009
Photo by Rod Bland | Agentic Systems Engineer (@rodbland)Photo by Rod Bland | Agentic Systems Engineer (@rodbland)Photo by Rod Bland | Agentic Systems Engineer (@rodbland)Photo by Rod Bland | Agentic Systems Engineer (@rodbland)Photo by Rod Bland | Agentic Systems Engineer (@rodbland)Photo by Rod Bland | Agentic Systems Engineer (@rodbland)Photo by Rod Bland | Agentic Systems Engineer (@rodbland)Photo by Rod Bland | Agentic Systems Engineer (@rodbland)Photo by Rod Bland | Agentic Systems Engineer (@rodbland)Photo by Rod Bland | Agentic Systems Engineer (@rodbland)
Replying to @TeslaSI
@teslasi @aelluswamy @TeslaAUNZ Loving FSD v14.3.9 on my Model Y L in Sydney. One thing to tune for left-hand traffic: it tracks well right of lane centre and gets uncomfortably close to large oncoming vehicles. Centring or a slight left bias would feel much safer.
1
2
115
Thanks folks at @AnthropicAI for giving me a laugh using Claude Code this morning 🤣
1
38
AI does have a sense of humor.
1
2
29
In case you are wondering about the typos, that's Windows Voice Access butchering what I said...
1
25
I shared a prompt called the Principal Architect Review for vibe-coded apps. Since then I found two follow-up prompts that take the results even further. Here's the full workflow 👇 🧵 THREAD 🧵
I use AI to build dashboards, write code, draft strategies. But AI ships its first draft. And first drafts have blind spots. So I built a prompt that forces AI to tear apart its own work before I review it. Here's the exact prompt (copy it): 🧵 THREAD 🧵
2
1
146
1/ After the PA review, bring in The Mentor: "I'm not so sure this is a thorough review, especially for customers and the business side. Now consider that a colleague who is a mentor of yours and a world-renowned UX expert came in and reviewed the application, its UI, how it works, its code quality. What would they say? Turn this into a sprint and execute on all the improvements."
1
1
16
My @openclaw agents frequently lost context and had no way to search across everything. Emails, Slack, Drive, other agents' conversations. All siloed. So I built Claw Recall, and just released v2.0.0. One searchable brain, updates in seconds, under $1/mo. Inspired by @natebjones's Open Brain. github.com/rodbland2021/claw…
1
1
39
I use AI to build dashboards, write code, draft strategies. But AI ships its first draft. And first drafts have blind spots. So I built a prompt that forces AI to tear apart its own work before I review it. Here's the exact prompt (copy it): 🧵 THREAD 🧵
1
1
1
192
1/ It's called the "Principal Architect Protocol." After your AI builds something, you say: "Now invoke the Principal Architect." It switches into ruthless review mode and rebuilds the thing. Here's each step:
1
1
13
This is exactly the kind of practical automation OpenClaw enables — no dashboards, just natural language control 🦞
Just ditched my smart home dashboard. Set up a local AI agent (OpenClaw) that connects to Home Assistant via the MCP protocol. Now I just text my house on Signal or Telegram to trigger physical devices. The "mcporter" skill handles the translation instantly. #HomeAssistant #AI
1
62
Same experience here. Stuff that used to sit on my todo list for weeks just... gets done now.
Used @openclaw + @AmpCode + @greptile to do a massive db migration to RLS. Took one day, one ralph loop with 154 tasks, and a huge PR, but it’s done. Would’ve easily been a month of work with a “normal” team pre-LLMs.
32
🧵 Built a monitoring dashboard for my @openclaw AI agent running on a VPS. Tracks gateway status, active sessions + token usage, CPU/RAM/disk, Docker, Tailscale, fail2ban, SSL certs, cron health, and alerts. Python stdlib + vanilla HTML/JS. 1/
3
4
291
Collector runs every minute via cron, writes a JSON file. Dashboard fetches it every 30s. No pip packages, no frameworks, no database. Requires Python 3.8+ and standard Linux tools. Docker/Tailscale/fail2ban optional. 2/
1
72
🧵 Built a fix for the @OpenClaw session corruption bug that was bricking all three model fallbacks simultaneously. Root cause: transcript repair inserts synthetic tool_result entries, but after compaction the matching tool_use_id no longer exists. Anthropic rejects the request → cooldown → infinite retry loop on ALL models. 1/
1
64
Klaivyo Email / SMS Marketing for Ecom - Ask The Experts with James Mayor and Rod Bland xitter.cf/i/broadcasts/1DXxyveoA…
206
Facebook video Ads tip: Focus on improving Outbound CTR, 3s video plays, Video video average play time and % 3s VV/Impressions. ROAS will then take care of itself.
Ok Facebook, what's up with YouTube video links not showing the thumbnails in Facebook organic posts? Conversely though, reach is 20x greater than usual?
1
Learning how to advertise successfully on Facebook is an acquired skill, like any other. It doesn't have to be complicated though. Here's a guide that keeps it simple: youtu.be/HxxeY0SLstg youtu.be/HxxeY0SLstg
If your ads aren't performing well, then maybe it's time to stop blaming the platform. youtu.be/-pq11Gbpne4