Jack Ngfinance × technology

Field note / killed

Masthead

A source-led system that turned current news into reviewed LinkedIn posts and carousel PDFs.

verdict: killed

#applied-ai#content#product#personal-brand

00 / The idea

Build a content operation that monitors relevant news, turns selected sources into a content plan, drafts LinkedIn posts and carousel PDFs in a defined voice, and publishes only after review.

01 / Why it’s appealing

  • It solved a problem I had myself: maintaining a professional voice without starting from a blank page each time.
  • Source-led drafts are more defensible than generic posts generated from an empty prompt.
  • The workflow joined discovery, planning, drafting, visual output, scheduling, and LinkedIn publishing in one system.
  • A user retained editorial control and could reject a draft rather than letting an autonomous system speak for them.

02 / What would have to be true

  • Professionals or firms must value a complete content operation enough to pay for it repeatedly.
  • The generated voice and analysis must be meaningfully better than combining alerts, a general chatbot, and a design tool.
  • LinkedIn access, publishing permissions, and platform rules must remain dependable.
  • Source ingestion and model use must produce reliable work at a cost that supports the service burden.
  • The product must solve the problem for users other than me without extensive setup and intervention.

03 / Risks

  • One committed user can create a detailed product without demonstrating a market.
  • AI writing and carousel generation are easy for larger platforms to absorb or commoditise.
  • Content quality, copyright, attribution, and brand risk still require human review.
  • Supporting many firms turns a personal workflow into a multi-tenant software and operations business.

04 / Verdict

killed

I built the product far beyond a prototype, but I could not establish a business model or a user need beyond my own.

Masthead monitored news, scored and classified sources, developed content plans, generated posts and carousel PDFs, scheduled drafts, and published through LinkedIn after approval. It eventually included onboarding, multiple brands, usage controls, billing infrastructure, and a substantial operating dashboard.

The system worked well enough to prove that it could be built. That was not the same as proving that it should become a business.

I needed it because I wanted a repeatable way to build a public voice. I did not have evidence that enough other people had the same problem, wanted the complete workflow, and would pay enough to support its ingestion, generation, and review costs.

I kept the lesson: automation can prepare the material and preserve the record, but the person whose name is attached still has to decide what is worth saying.