Everything here is live.

Problem

Freight brokers lose deals to slow manual re-entry. An incoming RFQ email triggers a cascade of copy-paste across three tools.

Approach

AI reads the email, extracts all fields, flags the ones that need human review. The broker confirms, not transcribes.

Outcome

Every field confirmed before the quote goes out. The tedious half of the job is gone.

Problem

Serious readers want the original language, the history and the scholarship behind a verse. Existing tools hide it behind one translation, or bury it in a format that takes graduate training to use.

Approach

Tap a verse, see what it meant. Interlinear text, commentary grounded in a corpus of historical scholarship, and a proprietary translation from a Claude pipeline that runs six integrity checks before a verse reaches the reader.

Outcome

Live on web, iOS and Android from one codebase, with all 1,189 chapters narrated. Every generated verse is logged and auditable, so a bad one can be fixed without rerunning the pipeline.

Problem

Every plant app gives you the name. None give you the story behind it, or a place to keep what you've found.

Approach

Identification card built backward from the result screen: etymology, origin, confidence score, two named alternatives. A Victorian botanical illustration generates async from the same photo.

Outcome

The prompt designed backward from the UI is what makes the AI output fit the product instead of the product fitting the API. The illustration isn't a feature. It's the reason you come back.

Problem

Every tool for self-expression opens on a blank public page. Most people never get past it, and the question of who they are stays unanswered.

Approach

Start private instead. Short voice, text, and photo captures. Claude reads all of them and builds a living portrait of the person's values, patterns, and story, in their own voice.

Outcome

Nothing goes public until the user says so. The most sensitive insight is locked private at three separate layers, so a bug in one can't leak it.

Problem

The real decisions happen in meetings, at the whiteboard and in Slack threads. By the time you open a task app, half the context is gone.

Approach

Capture first. Voice, photos and messages run through one Claude pipeline that pulls out the action items, decisions and owners, and flags what's missing.

Outcome

Action items arrive sorted by owner and by how sure the model is. You review the uncertain ones and act on the rest.

Problem

Project tools are built for enterprise teams with an ops person to run them. Startup teams of 2 to 15 just need everyone to know what's happening without a weekly status meeting.

Approach

AI does the update work instead of reminding you to: Claude summarizes threads you haven't opened in days, writes a daily digest of what needs you, and drafts the Monday check-in from real activity. Flat per-workspace pricing, never per seat.

Outcome

Live with paying clients. Every Barael client engagement runs on it.

Problem

Most AI PM interview prep ends at content. Candidates read and watch, then walk into the interview never having practiced under pressure.

Approach

Practice over reading. Claude writes a fresh PM scenario grounded in 200+ indexed Lenny's Podcast episodes, scores your spoken or typed answer with cited coaching, and spaced repetition brings each concept back.

Outcome

Live with paying students: 19 authored modules, plus an instructor platform that lets other creators run the same loop on their own content.