[{"data":1,"prerenderedAt":87},["ShallowReactive",2],{"post:deepseek-monitor-github-pull-requests":3,"posts":65},{"slug":4,"title":5,"date":6,"dateNice":7,"dateShort":8,"lastEdited":6,"excerpt":9,"tags":10,"readTime":14,"emoji":15,"cover":15,"body":16},"deepseek-monitor-github-pull-requests","I let DeepSeek read every pull request so I don't have to","2026-06-04","June 4, 2026","Jun 4","AI made writing code cheap, but reviewing it still eats my mornings. So I built a small tool that hands the first-pass pull request review to DeepSeek for about a dollar a day.",[11,12,13],"claude-code","tooling","notes",3,null,[17,20,22,25,28,30,32,34,36,38,42,44,46,48,50,52,54,56,58,59,61],{"type":18,"html":19},"p","Writing code got cheap. Reviewing it didn't. With Claude Code humming along, my team ships pull requests faster than ever — and every one of them lands on me to check. A pull request, or PR, is the proposal a developer submits before their code gets merged into the main project; someone has to read it and approve it. That someone was me, every morning, across half a dozen repositories. The coding got automated. The bottleneck just moved to my eyeballs.",{"type":18,"html":21},"This post covers a small tool I built to hand off the boring first pass of that review, what it costs, and why I'd rather pay a machine to do it.",{"type":23,"url":24},"image","\u002Fnotion-cache\u002F37ae2ae489132c9e.png",{"type":26,"html":27},"h2","What the tool does",{"type":18,"html":29},"It watches the repositories I care about and, whenever a new PR shows up, fires it off to DeepSeek V4 Pro — a low-cost AI model — for a first-pass review. DeepSeek reads the diff, runs a few different review angles over it (correctness, security, maintainability, test coverage), and reports back whether it found anything worth a human's attention.",{"type":18,"html":31},"Most PRs come back clean. The ones that don't get flagged, fixed, and only \u003Cem>then\u003C\u002Fem> land in front of me. By the time I look, the obvious stuff is already sorted. I review the genuinely interesting changes instead of rubber-stamping submodule bumps at 9am.",{"type":26,"html":33},"Why this is worth it",{"type":18,"html":35},"Two reasons: money and attention.",{"type":18,"html":37},"The money part is almost a joke. Each PR review runs a few cents — most under ten — and a full day of watching every repo comes in around a dollar, well under the cap I set.",{"type":39,"emoji":40,"html":41},"callout","💡","A typical review in my dashboard: roughly 5,000 tokens in, 2,400 out, about 26 seconds, costing under a cent. "Tokens" are the units AI tools bill by — think of them as fragments of words.",{"type":18,"html":43},"The attention part matters more. Code review is the kind of task that never ends — there's always another PR in the queue — and reading diffs all day is genuinely draining. Offloading the first pass means I spend my brain on business logic, architecture, and how the pieces fit together, instead of on the hundredth "did they remember the null check" of the week.",{"type":26,"html":45},"Running it yourself",{"type":18,"html":47},"It's a command-line tool — meaning you run it in a terminal, the text window where you type commands instead of clicking buttons — and it lives entirely on your own machine. It polls your repos, sends new PRs to DeepSeek, and prints a live dashboard of what's open, what's been reviewed, and what each review cost.",{"type":18,"html":49},"The code is on GitHub if you want to point it at your own repositories.",{"type":26,"html":51},"Verdict",{"type":18,"html":53},"Coding is cheap now. Human attention isn't. Spending a dollar a day to keep my mornings clear of routine review is the easiest trade I've made all year — the value isn't the dollar I save, it's the hour I get back.",{"type":55},"hr",{"type":18,"html":57},"Next up: what DeepSeek catches that I'd have missed, and what it misses that I catch.",{"type":55},{"type":26,"html":60},"References",{"type":62,"items":63},"ul",[64],"\u003Ca href=\"https:\u002F\u002Fgithub.com\u002FEdmundHee\u002Fcode-review-cli\" target=\"_blank\" rel=\"noopener noreferrer\">code-review-cli on GitHub\u003C\u002Fa>",[66,68,77],{"slug":4,"title":5,"date":6,"dateNice":7,"dateShort":8,"lastEdited":6,"excerpt":9,"tags":67,"readTime":14,"emoji":15,"cover":15},[11,12,13],{"slug":69,"title":70,"date":71,"dateNice":72,"dateShort":73,"lastEdited":71,"excerpt":74,"tags":75,"readTime":76,"emoji":15,"cover":15},"deepseek-v4-pro-vs-claude-code","DeepSeek V4 Pro: 519 million tokens for $7","2026-05-16","May 16, 2026","May 16","Our four-dev team blew through Claude Code's weekly budget, so we pointed everything at DeepSeek V4 Pro instead. Twelve days and $7 later, here's how it stacks up against Opus 4.7 and Sonnet 4.6 — and why the timing matters.",[11,12,13],4,{"slug":78,"title":79,"date":80,"dateNice":81,"dateShort":82,"lastEdited":71,"excerpt":83,"tags":84,"readTime":85,"emoji":86,"cover":15},"cutting-claude-code-tokens-with-caveman-mode","Cutting Claude Code tokens by ~75% with caveman mode","2026-05-14","May 14, 2026","May 14","An ultra-compressed prompt style that strips articles, filler, and pleasantries from Claude Code without losing technical substance. Here's what it costs, what it saves, and where it bites.",[11,12,13],6,"🪨",1780589404138]