[REDACTED] Episode 5: The Rage Log: AI-Powered Lead Gen, a Living Landing Page, and the Trick That Stops Claude From Making the Same Mistake Twice
How we're breaking down their AI lead gen pipeline, a B2B landing page built entirely with Claude, and the "gotcha registry" technique for baking past mistakes into every future build
We need your [Redacted] AI experiences for upcoming episodes! Who's the most underrated AI builder you know? Someone running real systems inside a real business? Send us a message at contact@tweenerfund.com because we want to get them on the show!
Episode 5
Redacted is the show that doesn't clean things up before hitting record. Episode 5 of Redacted goes deep inside Offline. The format is the usual one: we take a turn showing what we’ve actually been building this week, then Taylor closes with a story about turning a 36-hour AI rage spiral into something genuinely useful. It's candid, specific, and doesn't pull punches about what's still not working.
What We Cover
The outreach brief: Before any LLM writes an email, Taylor’s system assembles a context document for each account that includes connected contacts, communication history, open support tickets, similar partners, and the reason for reaching out. That brief becomes the input to the writing step, making outreach context-aware instead of generic.
B2B Hinge: Taylor’s lead graph is organized into three tiers: green for existing Offline partners, yellow for warm connections that are one or two degrees away, and blue for cold, in-market prospects. He describes it as “a B2B version of Hinge” because nearly every target account is only one or two introductions away.
2,000 locations geocoded in ~1 hour: Claude wrote a script that called the Google Places API to add latitude and longitude to every restaurant location in HubSpot. The process only produced around 10 errors that were fixed manually, turning what would have taken weeks into about an hour of work.
Replacing LLMs with code: Taylor has been intentionally replacing prompt-based workflows with deterministic code. For market geography, he swapped LLM prompts for latitude, longitude, and radius fields, making the system more reliable, easier to audit, and simpler to update without changing prompts.
The human-AI conveyor belt: Taylor’s automation process depends on a feedback loop where Steve first trailblazes the sales process, Taylor automates it, and Steve then surfaces edge cases one lead at a time. Instead of sharing broad observations, Steve walks through specific examples, allowing Taylor to improve the system based on concrete failures.
The rage log and gotcha registry: Taylor asked Claude to review every conversation where he became frustrated, identify recurring mistakes, and compile them into what Claude called a “rage log.” He refined those findings into a “gotcha registry” that documents known failure patterns before new work begins.
Hooks as memory engineering: Rather than relying on
claude.mdfiles or built-in memory, Taylor uses Claude Code hooks to automatically inject the gotcha registry into every planning step. As a result, every implementation plan is checked against previous failures before any code is written.The Slack bot PM experiment: David built a Claude agent that managed the final landing-page asset collection process by pinging teammates in Slack and checking for updates every 15 minutes using the
loopcommand. The experiment showed that autonomous agents work better when they have a clear manifest defining exactly what to check and what actions to take.AI photo tagging at scale: David used an inexpensive AI model to tag 2,000–3,000 images in Offline’s Google Drive with marketing-relevant metadata such as whether photos featured individuals, couples, groups, food, events, hospitality staff, or bartenders. This made it possible for Claude to retrieve images by category instead of requiring manual browsing.
The trust argument for context-aware outreach: Taylor argues that as AI-generated outreach becomes more common, the value of emails that demonstrate genuine knowledge of an account increases. His trust graph is built on the idea that AI dramatically lowers the cost of gathering context while making personalized, context-rich outreach a stronger competitive advantage. How to watch:
It’s best viewed on YouTube to fully see the examples (make sure to subscribe!)
But also available on all audio podcast players through Tweener Talks!
PLUS we have a new spot for show notes and files discussed in the episode. Check cit out: https://github.com/instanttaylor/redacted-podcast
What’s Next?
New episodes drop twice a month/every other Wednesday. If you want to be on the show as a guest and show your [REDACTED] builds, email us here.