AI Tool

Human, Actually

Stronger job applications, grounded in what you have actually done.

Try Human, Actually
Type
AI Tool
Platforms
Web
Status
Live
Role
Design + engineering

The story

Why it exists

A resume rarely captures a full career. When generic AI generates from that incomplete record, it tends to invent, exaggerate, or produce material a candidate cannot confidently defend.

My role

Solo product designer and engineer. I defined the product, designed the experience, built the application and AI workflows, secured it, tested it, and shipped it.

Product move

Establish what is true before generating anything. Build a traceable evidence base, expose fit and gaps, ask focused questions, then create role-specific materials.

What shipped

A free production application that analyzes opportunities, interviews candidates about missing context, and generates grounded resumes, cover letters, and application answers.

See the product

Targeted questions gather the missing evidence most likely to strengthen the application.
Tailored materials are generated only after the evidence base is established.
Candidates choose their provider while their credentials remain encrypted.

Product judgment

Key decisions

  1. 01

    Evidence before generation

    Every useful claim is grounded in a source or confirmed answer, with provenance that distinguishes direct, inferred, and unsupported evidence.

  2. 02

    Ask when information is missing

    A gap-based interview gathers specific context instead of letting the model invent a plausible substitute.

  3. 03

    Design each artifact separately

    Resumes, cover letters, and application answers have distinct workflows because they require different judgment, structure, tone, and review.

Build notes

Fit is visible at the requirement level rather than collapsed into one opaque score. The case state then recommends the next useful action: add a source, answer a question, refresh the analysis, or review an output.

The production application includes source ingestion, evidence extraction and provenance, adaptive interviews, natural-language chat, ATS-aware resume QA, cover-letter tone controls, per-question assistance, and public-presence auditing.

Long-running ingestion and analysis run as background jobs. Structured outputs keep AI responses predictable across OpenAI, Anthropic, and Gemini, while user-provided credentials are stored encrypted.

I owned the work from product thesis through deployment and iteration. Adam Cobb and Eli Yelluas supported penetration testing and security review.

Built withNext.js, TypeScript, PostgreSQL, Prisma, Structured AI outputs, Background jobs, OpenAI, Anthropic, Google Gemini, Encrypted credentials

Next product

WatchGPT

Realtime ChatGPT voice, on your wrist.

View product
WatchGPT product preview