[]
0%
Home Work Photos About Contact
Jake Davis

Software developer. Photographer. Neuroscience student. Creator.

I build AI systems and automation tools — agents that run autonomously, software that removes me from the loop.

Right now: a multi-agent system on a Mac Mini that manages my schedule, trades crypto and prediction markets, and ships code on its own. The commit history doesn't sleep.

I shoot photos. Mostly outdoors, mostly film.

Python · JavaScript · TypeScript · Swift · SQL · Bash · AppleScript
Astro · React · Node.js · SQLite · Docker · Git
Claude API · OpenAI API · Canvas API · Brave Search API · ElevenLabs API · Telegram API · Polymarket API · Tailscale · Linux · macOS
AI/ML · Agent Architecture · Automation · Web Dev · Systems Design · Market Analysis
yerb — mac mini
> auto-accept Yerb 2.8
0 tokens 0%
yoursite.com
Sign up
● Now in beta
Ship faster.
Build smarter.

The modern platform for teams who want to move fast without breaking things.

Get started View docs →
99.9% Uptime
2.4M Requests/day
<50ms Avg latency
100 Performance
100 Accessibility
100 SEO
0.4s Load
9:41
Today Mar 4
1,847 kcal
of 2,400 goal 553 remaining
Protein 142g
Carbs 186g
Fat 61g
Meals 3 logged
Breakfast Eggs, toast, avocado
520
Lunch Chicken bowl, rice
785
Snack Protein shake
542
CI / CD ● main
Build
23s
Test
1m 47s
Lint
12s
Deploy
...
Test Suites
auth.test.ts 24 passed
api.test.ts 61 passed
db.test.ts 38 passed
utils.test.ts 24 passed
147 passed · 0 failed
Commits
feat: add rate limiting a3f2c91 · 4m ago
fix: token refresh race e7b1d04 · 22m ago
refactor: extract middleware 1c8e3fa · 1h ago
merge: feature/auth 9d4a2b7 · 3h ago
Claude
Build me a dashboard
I'll set up React with D3 for the charts. Adding real-time data via WebSocket and a responsive grid layout.
Add dark mode too
POST /v1/chat/completions 200 stream
{
  "model": "claude-sonnet-4-20250514",
  "max_tokens": 1024,
  "messages": [{
    "role": "user",
    "content": "Analyze this dataset"
  }]
}
I'll start by loading the CSV and checking for null values across all columns, then compute summary statistics
~/projects
$ npm create astro@latest
  ◼ tmpl   Using: blog
  ◼ deps   Installing...
  Done
$ npm run build
  building (SSG)...
  14 pages built in 1.86s
$
agent.py
 1  import anthropic
 2
 3  async def run(prompt: str) -> str:
 4      client = anthropic.Anthropic()
 5      msg = client.messages.create(
 6          model="claude-sonnet-4-20250514",
 7          max_tokens=1024,
 8          messages=[{
 9              "role": "user",
10              "content": prompt
11          }]
12      )
13      return msg.content[0].text
EDGE GATEWAY SERVICES DATA client:443 TLS nginx:80 LB node:3000 auth node:3001 api node:3002 workers postgres:5432 redis:6379 2.4k req/s
● Active
Scrape listings
Score deals
Send alerts
[14:02:31] Fetched 47 new listings
[14:02:33] Match: 2019 Tacoma — $28,400
[14:02:34] Scoring batch 2/3...
[14:02:35] _
GET /api/v1/projects 200 23ms
{
  "data": [
    {
      "id": 1,
      "name": "fuel-ios",
      "status": "live",
      "stars": 128
    }
  ],
  "total": 6,
  "cached": true
}

I'll build
your site

I design and develop fast, custom software — no templates, no bloat. Hand-coded with modern tools, deployed in days not months. If you want something that works and ships fast, let's talk.

01
Custom Built No templates. No boilerplate. Built to spec from scratch.
02
Fast Sub-second load times. Lighthouse 100s across the board.
03
Right Tool, Right Job Swift, Python, React, Astro — whatever fits the project.
Start a project →

Projects

(hover)
01

Fuel

Native iOS calorie tracker with AI food parsing, barcode scanning, HealthKit sync, and Apple Watch companion.

Status Live
784
left
P 86g
C 134g
F 42g

Type what you ate in plain English — AI parses it into calories, protein, carbs, and fat with confidence scores. Barcode scanning for packaged foods. HealthKit sync, Apple Watch companion, and home screen widget.

Stack Swift · SwiftUI · SwiftData
Input AI parse · Barcode · Manual
Sync HealthKit · Watch · Widget
02

Yerb

Personal AI assistant running 24/7 on a Mac Mini. Manages calendar, automates coursework, runs autonomous builds overnight.

Status Active
yerb
scheduler
market
scraper
builder
847h uptime · 4 agents
>

24/7 autonomous AI assistant on a Mac Mini. Manages calendar, automates coursework, runs overnight builds, hunts vehicle deals, and drafts content — all without being asked.

Runtime OpenClaw · Claude · Python
Agents Scheduler · Market · Scraper · Builder
Infra Mac Mini · Always-on · launchd
03

Token Tracker

Native macOS widget tracking Claude Code usage in real-time. Parses sessions, shows per-model breakdowns and costs.

Status Active
* Claude Token Tracker
14.2M
lifetime tokens
Today 847.3K
Opus ~$142.80
7-day usage
M
T
W
T
F
S
S
Opus 8.4M
Sonnet 4.1M
Haiku 1.7M

Native macOS widget that parses Claude Code session files in real-time. Tracks per-model token breakdowns across Opus, Sonnet, and Haiku with estimated cost calculations. Multi-device iCloud encrypted syncing, Discord rich presence support. Reads directly from ~/.claude — fully local, fully private.

Stack Swift · SwiftUI · WidgetKit
Updates Every 15 min via timeline provider
Sizes Small + Medium widget families
04

Marketplace Scraper

Automated vehicle deal finder for Toyota Tacoma & 4Runner. Scrapes listings, scores deals, sends alerts.

Status Active
DEAL ALERT 87
2016 4Runner SR5
$24,500
82k mi · 4WD · Clean title
Price
Miles
Year
via Telegram

Automated Craigslist scraper that finds Toyota Tacoma and 4Runner deals. Scores listings by price, mileage, year, and condition — sends high-scoring matches straight to Telegram.

Stack Python · BeautifulSoup · SQLite
Alerts Telegram Bot API · Instant push
Schedule Cron · Integrated into Yerb
05

Neural Networks

Custom neural network library — backpropagation, activation functions, convolutional layers — no frameworks.

Status In Progress
forward pass
inputhiddenhiddenoutput

Neural network library built from scratch with only Python and NumPy. Dense layers, conv layers, backpropagation, activation functions — every gradient computed by hand.

Stack Python · NumPy · Matplotlib
Layers Dense · Conv2D · Activations
Training SGD · Momentum · LR scheduling
06

Vault Keeper

AI-powered accountability system built on Obsidian. Reads journals, detects behavioral patterns, generates briefs.

Status Legacy
Weekly Brief
Commitments kept
68%
Gym consistency
40%
Deep work hours
82%
Pattern: overcommit Sun → burnout Tue
LEGACY → Yerb

AI accountability system built on Obsidian. Reads journal entries, detects behavioral patterns, and generates weekly briefs using your own words against you. Evolved into Yerb.

Stack Python · Claude API · Obsidian
Analysis Pattern detection · Self-sabotage tracking
Output Friday briefs · Recovery plans

Get in touch

For projects, collaborations, or just to say hey.