How I AIThe AI content machine that turns ideas into posts that don't sound like slop | Alex Lieberman
CHAPTERS
- 0:00 – 2:24
Why content is a leverage problem: time limits + employee creator enablement
Alex frames the “content machine” as a solution to two bottlenecks: his personal time cap for creating content and the difficulty of getting busy employees to create consistently. The goal is an AI-native/AI-assisted workflow that increases output without producing generic, voice-less posts.
- •Content has created major career opportunities, but Alex is time-constrained
- •Re-engineering the content process to be AI-assisted while avoiding “AI slop”
- •Building a media layer on top of a non-media business as a moat (trusted distribution)
- •Making it frictionless for employees to become creators despite full-time roles
- 2:24 – 6:57
Distribution beats “build it and they will come”: Tenex context and the media-company thesis
Alex and Claire discuss why distribution matters more than ever, especially as technology becomes commoditized. They connect employee-led content to trust, hiring, and long-term defensibility.
- •Tenex as an applied AI company helping enterprises through AI transformation
- •In a post-AI world, product moats shrink; distribution and trust become critical
- •Turning employees into creators is underused but powerful
- •Claire’s exec perspective: buyers often “buy the team,” not just the product roadmap
- 6:57 – 9:15
What “AI slop” really is—and where AI should/shouldn’t write
They tackle the core objection: LLMs often write bland copy. Alex argues the biggest risk isn’t AI drafting per se, but weak thinking and weak inputs; AI can help in many steps without capping quality.
- •Nuanced take on ‘AI raises the floor but caps the ceiling’ for elite writers
- •AI can raise the ceiling in idea discovery/research even if drafting may average out
- •For teams, the real alternative is often ‘no content at all’
- •Reframing slop: poor ideas and vague inputs create poor outputs
- 9:15 – 13:28
Mapping the end-to-end content workflow before adding AI
Alex breaks content creation into concrete steps (idea → research → thoughts → format → draft → edit → publish → repurpose). Claire reinforces the broader lesson: workflow-first AI adoption works beyond marketing.
- •Canonical workflow: inspiration, research, brain-dump, choose format, write, review/edit, publish, repurpose
- •Deciding whether AI is driver vs copilot at each step
- •Design the “ideal world” workflow first, not today’s constrained version
- •Process mapping often reveals waste even before adding AI
- 13:28 – 13:58
Rebuilding from scratch: the Content Machine concept and system connections
Alex introduces the Content Machine as a repo of modular ‘skills’ that can run via CLI/Claude Code/Cowork and connect to company systems. The system is designed to pull from real work artifacts to generate grounded content ideas.
- •Content Machine as a directory of chained skills living in a repo
- •Connects to systems of record: Slack, Notion, meeting notes, Linear, Git, Gmail
- •Built to be usable across a company, not just by Alex
- •Foundation: use internal data + external reading to fuel idea generation
- 13:58 – 16:21
Step 1 — The Oracle: weekly scanning + scoring to generate “content spikes”
The Oracle scans the last seven days across internal tools and curated internet sources, then ranks potential post ideas with a scoring rubric. Alex calls this the single most helpful piece because it eliminates blank-page paralysis.
- •Daily list of ~15 ranked “content spikes”
- •Scoring signals: anecdotes, strong POV, specificity/examples
- •Half internal (work artifacts), half external (internet reader of followed accounts/sites)
- •Primary value: faster jump from blank page to concrete idea
- 16:21 – 16:36
Step 2 — Research assistant (optional): briefs, angles, and open questions
If Alex chooses a spike he doesn’t fully understand, he can run a research step that produces a structured brief. The output is meant to support better interviews and more grounded claims, not to invent a generic take.
- •Research brief summarizes the topic and key viewpoints
- •Suggests contrarian/novel angles and open questions to answer
- •Lets Alex pull quotes/facts during the interview stage
- •Used selectively when the creator lacks full context
- 16:36 – 16:51
Step 3 — Interview Panel: extracting specificity with interviewer personas + voice input
Alex uses an Interview Panel made of well-known interviewer personas to pull detailed stories and sharp thinking. He answers via ‘yap-to-text,’ producing a transcript that becomes the primary source material for the draft.
- •Panel personas (e.g., Tim Ferriss, Joe Rogan, Barbara Walters, Howard Stern, etc.)
- •Goal: force specificity, examples, and clear reasoning
- •Voice capture via Wispr Flow (spoken answers → transcript)
- •Transcript becomes the core content; AI shapes rather than invents
- 16:51 – 20:24
Personalization layer: style guide, voice guide, and the ‘content lessons’ memory loop
Alex shows how the system avoids generic writing by grounding drafts in a personal style guide and a voice guide learned from his best-performing posts. A ‘lessons’ file accumulates recurring feedback so the system improves over time.
- •Style guide: role context, assets to promote, preferred formats, integrations, followed accounts
- •Voice guide: tone, hook formulas, structures, patterns from top posts; “text a friend” rule
- •‘Content lessons’ Markdown logs recurring mistakes (tone, structure, phrasing)
- •Lessons loop compares diff between draft and final edit, proposes reusable rules, and saves them
- 20:24 – 23:21
Step 4–6 — Drafting + Writer’s Council: scoring, revision loops, and repurposing
After the interview transcript is created, the system drafts a post in Alex’s voice and runs a Writer’s Council of writing personas that score the piece. If it’s below threshold, it iterates until it reaches the target score, then supports repurposing into multiple formats.
- •Draft uses voice/style guides + lessons file; transcript is the main source of wording
- •Writer’s Council personas (e.g., David Perell, Morgan Housel, etc.) score 1–10
- •If aggregate < 9/10, system runs revision loops until it passes
- •Repurposing step converts an anchor post into tweets/LinkedIn variants in prior proven formats
- 23:21 – 31:14
Live demo in Claude: Oracle results → research brief → interview → draft critique
Alex walks through a real run of the system inside Claude: pulling spikes from internal notes and his internet reader, selecting an FDE debate topic, generating a research brief, and producing a LinkedIn draft. They discuss how “slop” shows up (cheesy hooks, generic phrasing) and how feedback becomes new lessons.
- •Example spikes include internal anecdotes and external debates (e.g., FDEs)
- •Research brief compiles arguments and suggests angles/questions
- •Interview panel asks targeted follow-ups to force concrete customer stories
- •Draft review: identify ‘AI-cringe’ patterns and add them to lessons; run Writer’s Council next
- 31:14 – 36:57
Employee advocacy as a behavior system: Tenex Creator Cup mechanics and goals
Alex explains the non-AI side: incentives, visibility, and internal culture to make employee posting consistent. The Creator Cup uses points, engagement rewards, editor’s picks, and team-based unlocks to keep it fun and accessible, with hiring and awareness as primary ROI.
- •Prior experiment (‘Own the Internet’) drove 40% of inbound leads in a quarter
- •Tenex Creator Cup: points for posting and engaging; Slack channel to surface posts
- •Weekly editor’s pick bonus + weekly and monthly games; $5K prize pool
- •Bootstrapped ‘underdog’ distribution strategy; employee platforms also help retention and recruiting
- 36:57 – 42:58
Lightning round: great engineers, personal AI use cases, and fixing slop outputs
They close with quick hits on what makes engineers great now, Alex’s favorite personal AI automations, and what he does when the system produces poor drafts. The emphasis returns to foundations + adaptability, plus being willing to revert to manual work when ROI drops.
- •Great engineers: deep systems understanding + willingness to go agentic
- •Forward-deployed engineering importance: business context drives AI transformation
- •Personal AI: curated job-matching email + auto-fill applications; custom children’s books
- •When outputs are bad: either write by hand or give blunt, detailed feedback to update lessons
