AI product · Publishing · 2026
Agentic Brief
I built an AI publishing product that combines the topics a reader chooses with the LinkedIn profile details they authorize, so each brief can follow their actual interests.
- Product
- A preference-led publishing signup combining chosen topics with user-authorised LinkedIn identity.
- Delivery scope
- Onboarding design, OIDC, signed form tokens, domain checks, abuse controls, privacy-safe logs, and workflow delivery.
- Public evidence
- 1owned product
Project constraints
- 01Useful preference data without a long form
- 02Identity data only with user authorisation
- 03Bot and repeated-abuse resistance
The challenge
Collect useful reader intent without creating a spam-prone anonymous form or a generic one-size-fits-all newsletter.
An email address alone cannot support a personal brief, but a long signup would push readers away. The form asks for topics in one pass, confirms identity through LinkedIn, and gives the publishing workflow a clear record of what each person wants.
Architecture and decisions
Topics captured clearly
The form saves each reader’s chosen topics in a way the publishing workflow can use, instead of hiding useful context inside one long note.
Safer signup
Layered checks reject malformed emails, oversized requests, bots, and repeated abuse before a signup reaches delivery.
LinkedIn profile enrichment
With the reader’s permission, LinkedIn supplies the profile details available through OpenID Connect. A signed, short-lived token ties those details to the correct form and rejects expired or tampered submissions.
Verified delivery
Only readers who complete LinkedIn verification enter the publishing workflow. Privacy-safe logs show failed attempts and retries without exposing raw email addresses.
My role
-
01
Designed the onboarding around what the reader wants to follow, not around a generic email capture.
-
02
Built the server-side validation, mail-domain checks, honeypot, body limit, and request-rate controls.
-
03
Implemented LinkedIn OpenID Connect and mapped the available, user-authorized profile fields into signed, expiring, form-specific tokens.
-
04
Connected each authenticated reader’s interests to the publishing workflow while keeping the public product easy to find and understand.
What changed
A signup that knows what each reader wants and verifies who is asking, instead of collecting anonymous email addresses.
The finished signup captures who the reader is and what they want without asking them to complete a long profile. It saves their interests in a form the publishing workflow can use, requires LinkedIn authentication before storage, and records failures cleanly. Agentic Brief can now prepare genuinely preference-led editions without a second email-confirmation step.
- Only readers who complete LinkedIn verification enter the active publishing workflow.
- Each signup includes chosen topics and the profile details LinkedIn shares with the reader’s permission, not just an isolated email address.
- Rate limits, Turnstile, email checks, and privacy-safe logs keep the signup reliable enough to run as a real product.
Tools and technologies
Sources
Agentic Brief is an owned product on this site. The working signup is available below.
Agentic Brief / Personal technology brief
A technology brief that gets narrower, not louder.
Each edition follows the topics you care about, from agentic products and developer tools to AI search and Bitcoin infrastructure.