Replace the content string at :25, :33 (og:description) and :46 (twitter:description) with: 203 applications. 76 rejections. 34 ghosts. 2,710 GitHub contributions. 6 products. A cross-section of May 2025 to March 2026, searching and building with AI.

A career cross-section · 10 months

Growth
Rings

May 2025 — March 2026. A record of 203 applications, 68 companies, 2 offers, and the 2,710-contribution build season that grew alongside the search.

Drought, fire, abundance, resilience. You can't fake a tree ring.

searchrejection / ghostingbuildingoffer

The Funnel

Conversion Funnel
Outcome Distribution
The Dual Track — Applications vs. GitHub Commits Nov '25: Claude Code

Where the Effort Went

Top 15 Companies by Applications
Effort vs. Outcome Key Insight
Volume didn't predict success. 5 single-application companies yielded interviews — the same number as all multi-application companies combined.
Single-app companies 35 companies
  → Got interviews 5 (14.3%)
Multi-app companies (2+) 33 companies
  → Got interviews 5 (15.2%)
The takeaway: Applying to 50 roles at one company produced the same interview rate as applying once.
What if the hours spent on apps 2-50 went to building instead? After November, they did.

Company Tier Breakdown

FAANG+
40%
interview rate
5 companies · 64 apps
Big Tech
8%
interview rate
12 companies · 41 apps
AI Frontier
20%
interview rate
5 companies · 24 apps
Mid-Market
13%
interview rate
46 companies · 74 apps
Tier Performance Analysis
FAANG+ Detail
Microsoft Offer 50 apps
Google Interview 7 apps
Amazon Rejected 1 app
Meta Ghosted 3 apps
NVIDIA Ghosted 3 apps
FAANG+ had the highest interview rate at 40% — but it's skewed by Microsoft's 50-app persistence campaign. Without Microsoft, FAANG+ interview rate drops to 25% (Google only).

The Ghost Problem

Companies that never responded Applications sent
The Ghost Economy
Half of all companies contacted — 34 out of 68 — never responded at all. Not a rejection, not a "we'll keep your resume on file." Nothing.
Ghosted
34 companies
Rejected
23 companies
Interviewed
8 companies
Offer
2 companies
Referral
1 company
62 applications (30.5% of all apps) went to companies that never responded. OpenAI and Atlassian absorbed 5 each.
The ghost months were the build months. November through March: 1,525 commits, 4 products launched.

What Actually Worked

Referral vs. Cold Apply +36.4pts
Referral
50%
interview rate (1 of 2)
Cold Apply
13.6%
interview rate (9 of 66)
Referrals converted at 3.7x the rate of cold applications. Two referral connections (Microsoft, Home Depot) generated 58 applications — 28.6% of all activity.
The Microsoft Persistence Story Case Study
50 applications. 30 rejections. 2 interviews. 1 offer.
Applied
50 roles
Rejected
30 rejections
Interviewed
2
Offer
1
24.6% of all applications went to one company. The offer didn't come from the ATS — it came from direct outreach.
The portfolio helped. By the time I made that direct outreach, I had a patent-pending mobile app, an AI SaaS product, a community platform, a gaming engine, and a workshop tool — all built with AI.
LinkedIn Easy Apply vs. Direct/ATS
LinkedIn Easy Apply
14.3%
interview rate (1 of 7)
Direct / ATS
14.8%
interview rate (9 of 61)
Nearly identical conversion. The channel didn't move the number.
The Cost of Each Outcome
Applications per interview20.3
Applications per offer101.5
Companies per interview6.8
Companies per offer34
Interview → Offer rate20%
Total rejection emails76
Once past the resume screen, 1 in 5 interviews converted to an offer. The hardest part wasn't the interview — it was getting in the room.

The Numbers Behind the Rejection Emails

Who Rejected the Most
Role Diversity — Companies Where Multiple Roles Were Tried

Per-Company Breakdown

Company Status Applications Rejections Interviews Referrals Emails

The Builder Narrative

While Job Searching, I Was Shipping GitHub
In November 2025 — month 6 of unemployment — I picked up Claude Code via a $1,000 API credit. I'd never made a single GitHub contribution before. In the 5 months since, I went from zero to 2,710 contributions across 33 repositories. While 34 companies ghosted me and 76 rejection emails piled up, I was shipping real products — an AI SaaS, a community platform, a patent-pending mobile app, a gaming engine, and a workshop tool — all with custom domains, all built with AI.
Contributions
2,710
from literal zero
Repositories
33
created in 5 months
Peak Month
559
commits in Jan '26
Pull Requests
1,141
merged across projects
Monthly GitHub Breakdown — Commits vs. PRs
Commits Per Day Consistency
4.6
Nov
10
Dec
18
Jan
10
Feb
10.8
Mar*
Average across all 5 months10.4/day
First-ever GitHub contributionNov 14, 2025
Days from zero to 33 repos~130
*Mar through 3/22
The Ramp — New Repos Created Per Month
What I Shipped Products
Products live as of March 2026 — in order of launch
brokenbranch.devPortfolio
AI platforms, products & teams. Built entirely with Claude.
scoutsplus.orgCommunity Platform
Adult merit badge community. 8 skill categories, local troops, physical enamel pins & patches, leaderboards. Full social platform with feed.
visualinventory.aiAI SaaS
AI-powered home inventory. Snap a shelf photo and Claude Vision identifies the items. Expiration alerts, receipt scanning, voice updates. Android early access.
RoadTripper.aiMobile App · Patent Pending · now thelongway.ai
AI-narrated road trips. 3 raccoon characters (Scout, Skates, Macey) narrate every mile. Spotify integration, auto spooky mode at sunset via GPS, family mode. iOS & Android.
exp-lore.aiDesktop + Cloud
A significance engine for gamers. Claude Vision watches gameplay, writes literary chronicles. Behavioral fact database, salience selection, cross-game retrospectives.
vibecrafting.aiAI Workshop Tool
AI workshop companion. Describe a build → cut lists, shopping lists, exploded views, 3D-printable connectors that replace joinery, OpenSCAD export.
Plus games, demos, and research tools. Full portfolio → brokenbranch.dev

The Toolkit

What I Built With Stack
Every product was built with Claude Code as the primary development partner. I didn't learn to code in the traditional sense — I learned to direct code. The PM skills (requirements, architecture thinking, user stories, edge-case paranoia) turned out to be the exact skillset AI-assisted development rewards.
Claude Code
AI pair programmer — zero GitHub contributions to 33 repos in five months.
Next.js / React
Frontend framework for all web products. App Router, Server Components, the works.
Supabase
Auth, database, storage, edge functions. The backend for Scouts+, Visual Inventory, and more.
Vercel
Hosting & deployment for every web product. Git push → live in seconds.
React Native / Expo
Mobile framework for RoadTripper.ai and Visual Inventory's Android app.
Tailwind CSS
Design system backbone. Every UI, every product, consistent design language.
TypeScript
Language of choice across the entire portfolio. Type safety for AI-generated code was non-negotiable.
Stripe
Payments & subscriptions for SaaS products. Checkout, webhooks, subscription management.
Electron
Desktop app framework for Exp-lore.ai's screen capture and gameplay analysis engine.
What actually transferred: none of these tools required a CS degree. They required knowing what users need and where the complexity budget should go. Those are PM instincts. Claude handled the syntax.

Growing an Audience

LinkedIn as a Build Log Distribution
While most job seekers use LinkedIn to apply, I started using it to publish. Two longform posts about the AI development experiment — "The $1K Claude Code Credit" and "If You Can Read a Recipe, You Can Now Be a Developer" — turned the job search narrative from "looking for work" into "building in public."
Published articles4
PCF whitepaperPublished on-site
The articles did more for visibility than 203 applications. People who read them reached out. The Microsoft offer came from someone who saw the portfolio, not someone who saw a resume in an ATS.
Products as Proof of Work Signal
Each shipped product gave me something to point at. Instead of "I can manage AI products," it was "here's one I built, here's the live URL, here's the GitHub."
Custom Domains
6
All .ai or .dev or .org
Patent Filed
1
RoadTripper.ai
Platforms Covered
4
Set the count at :1269 to 3 and this line to:
Web, Android, Desktop
Product Categories
6
SaaS, Community, Mobile, Gaming, Workshop, Portfolio

What the Rejections Freed Up

Rejection as Raw Material Growth
Every "we've decided to move forward with other candidates" email freed up a block of attention. Instead of tracking application status, I built an AI SaaS product and filed a patent. The constraint — no income, shrinking runway, mounting pressure — forced ruthless prioritization. Only ideas that could ship fast survived. Only features users actually needed made the cut.
May – Oct
The Grind
Six months of applying. Optimizing resumes, tailoring cover letters, tracking in spreadsheets. Diminishing returns. The ATS void.
Nov – Dec
The Pivot
Claude Code arrives. First commit Nov 14. 15 repos in month one. The energy that went into applications redirected into products. Still applying, but now also building.
Jan – Mar
The Compound
Peak output: 559 commits in January alone. 13 more repos across Jan–Mar. The portfolio is the resume. Two offers arrive.
The months with the most rejections (Jan: 35 apps) were also the most productive build months (Jan: 559 commits).

What I Learned

10 Rings, 10 Lessons Dendrochronology
Ring 01
The ATS is a black hole
50% of companies ghosted completely. Treat applications as lottery tickets, not conversations.
Ring 02
Volume is a vanity metric
Single-app and multi-app companies had identical interview rates (~14-15%). The first application is the signal. After that, you're just generating noise.
Ring 03
Referrals converted better — on two data points
50% interview rate versus 13.6% cold — on a base of two referrals and 66 cold applications.
Ring 04
The hardest part is getting in the room
Once past the resume screen, 1 in 5 interviews converted to an offer: 2 offers from the 10 companies that got me into a room.
Ring 05
Build in public, not in silence
LinkedIn articles about the AI development experiment generated more warm interest than 200+ applications through official channels.
Ring 06
No bootcamp, no certification
I didn't take a bootcamp or certification course. I shipped 6 products instead.
Ring 07
AI-assisted dev rewards PM instincts
Requirements clarity, edge-case thinking, architecture decisions, user empathy — these are the skills that make Claude Code productive. Syntax is the easy part.
Ring 08
The channel doesn't matter
LinkedIn Easy Apply: 14.3% (1 of 7). Direct/ATS: 14.8% (9 of 61). Identical. Stop optimizing the envelope.
Ring 09
Constraint breeds invention
No income, no team, no design resources. The constraint forced fast iteration and ruthless scope.
Ring 10
Where the offer actually came from
The Microsoft offer came from direct outreach backed by a live portfolio — not from application #50 through the ATS. Build the proof, then show the people who matter.
The Full Picture Thesis
203 applications. 76 rejections. 34 ghosts. 2,710 GitHub contributions. 33 repos. 6 live products.

In month 6 I picked up Claude Code and started directing builds: an AI-powered SaaS product, an adult merit badge community, a patent-pending AI road trip companion, a gaming significance engine powered by Claude Vision, an AI workshop tool that turns a description into cut lists, and a developer portfolio — all from scratch, all while the inbox stayed cold.

The offer didn't come from volume. It came from direct outreach.