đŸȘšMason Thompson—Product Manager / Builder—SF Bay
Product Systems·

Console — Job Search Automation

Core idea

Job hunting is a numbers game. Automate the tedium of discovery, tracking, autofill, and status monitoring, and focus your attention on what matters most: creating high-intent applications.

Console covers similar territory to Simplify Copilot, but I built it as a self-contained system around my own workflow, local inference, and post-application observability for a couple of reasons. One, I don’t like paying for things I can build myself. Two, this material is fun and provides useful experience building with generative AI.

Optional, but I also have some thoughts on using generative AI in job applications — .

Restraint

I’ve kept generative AI use minimal, reflecting my belief that I need to understand and trust my local LLM before I can responsibly delegate work to it. That’s a high bar, especially in a hiring environment that Greenhouse CEO Daniel Chait called an “AI doom loop”.

My ultimate goal is to get a job. Console helps me apply ethically while capturing most of the practical gains from automation. But a very close second is understanding generative AI well enough to defend how the system works: learning from where it fails, deciding what should remain human, and being able to explain exactly what it is and isn’t allowed to do.

01 Feed

Aggregates live job postings from more than 18,000 companies across major ATS platforms, including Greenhouse, Ashby, Lever, BambooHR, Breezy, Teamtailor, Workday, and Workable, into a single searchable feed.

02 Apps

Tracks each application using multiple independent signals rather than relying on a single status field:

  • Posting state
  • Application history
  • Rejection emails detected through Gmail
  • RĂ©sumĂ© engagement detected through signed attribution links

Console also supports probes for informal outreach that is not tied to a formal application.

03 Résumé

Keeps the base résumé static. For each role, Console:

  • Extracts relevant keywords from the job description
  • Interprets those keywords in the context of the role
  • Maps them against evidence in my existing rĂ©sumĂ©
  • Generates only a tailored “Why I’m applying” section using a local LLM

The model can reframe existing experience and connect it to the role, but it cannot invent qualifications.

Example

“I’m applying because {COMPANY_NAME} Product Builder role sits directly at the intersection of my experience leading API and platform products, building hands-on prototypes, and developing AI workflows.”

Other features

  • Reusable autofill: previously entered answers can be reused across applications, including standard voluntary demographic fields. These values are explicitly supplied by the user and are never inferred or generated.
  • RĂ©sumĂ© attribution: signed links can trigger a push notification when an employer opens their rĂ©sumĂ©, providing a small amount of visibility into an otherwise opaque hiring process.
  • Local LLM agent — ama_Car: a parody of Jensen Huang’s statement on Dwarkesh, “we are not a car.” I carried over Jensen’s assets from ALL-LLLM Pod and turned him into ama_Car, where he exists as a head in a bubble. Like Microsoft’s Clippy, he offers occasional helpful remarks and uses synthetic speech to call attention to useful information.