THOL

• 

  Setup Settings

• 

  Where posts are generated

Inspiration

As a two-time founder, I so much time deciding what to post. THOL fixes that. Set your platforms (LinkedIn, TikTok, IG), add accounts you admire, set a goal, drop in ideas, and wake up to drafts ready to review. Schedulers already exist. What's missing is something that writes in your voice and still leaves the final call to you.

What it does

Onboard once: voice, pillars, platforms, inspiration accounts, seeds. Then every morning you get a review queue. Edit, approve, save, or mark as posted. Manage seeds and settings whenever. v1 is manual publish only. THOL never posts without you.

How we built it

We built it around a real morning habit, not a feature list. The rule came first: prepare overnight, human reviews, no autopost in v1. We modeled drafts, seeds, sources, and token limits in Postgres so everything had a home before AI touched it. Then a Next.js app with a Review / Seeds / Settings flow, Supabase magic-link auth, and onboarding that actually saves your voice and platforms. Draft generation is a server route: it pulls your settings, one seed, and ranked signals, calls Claude, and writes an in_review draft. Secrets stay on the server. When the first build looked like every other AI dashboard, we scrapped it and rebuilt it as a quiet morning desk, so the queue is the product, not the UI around it.

Challenges we ran into

The scope was massive: auto-scheduling, inspiration analytics, trends, TikTok image editing, learning from stats. We cut it to review-first. Trends and regional LinkedIn feeds are too unpredictable to trust with auto-posting, so we didn't. Earning trust mattered more than clever output.

Accomplishments that we're proud of

One clear rule that holds: THOL prepares, you decide. Auth and onboarding that persist to Supabase. A Claude draft API writing real drafts into the queue. A schema ready for sources and future learning. A focused UI built around one workflow: the morning.

What we learned

Generating text is easy. Voice and trust are hard. A ready morning queue beats another scheduler. A few finished screens beat five half-built ones. LLMs do better on compact signals than raw scrapes.

What's next for THOL

Make THOL the tool experts reach for when they want a consistent presence without living in a blank compose box. Next up: Apify inspiration intake, nightly draft jobs, Brevo morning email, performance learning, and IG share-from-TikTok. Scheduling and export come later, once manual publish feels rock solid.

Built With

• anthropic
• apify
• brevo
• nextjs
• tailwind