I Was Laid Off With 60 Days of Savings. I Built a Job Search Pipeline With Kimi K3.
The Situation
August 12th, my entire team got called into a 15-minute meeting, and that was that — 2 years at a mid-size SaaS company, 8 months of savings, two kids. I'm not going to pretend this is a productivity story. This is a survival story with a workflow in it. My previous job search, three years ago, was 120 blind applications, 4 callbacks, and a lot of lying to myself that quantity was the strategy. This time I had a real constraint: results in weeks, not months.
I'd been using Kimi K3 for dev work. On day three of the search I wrote a direction doc for myself — "what do I actually want and what do I actually have" — and realized the search needed a pipeline the same way code needs a CI. Here's the entire thing, unchanged.
The Pipeline
Every job posting goes through the same five steps, all in one K3 session per posting, with the same skeleton prompt I refined over the first week:
- JD dissection. Paste the posting, my direction doc, and my master resume. Ask for: the 8-12 requirements ranked by how central they are to the role, and — the useful part — which of them my resume already proves, which it implies, and which are missing entirely. That third list decided whether to apply at all. I stopped applying to jobs with 3+ hard gaps and my callback rate did the work.
- Resume tailoring. Not a rewrite — a re-emphasis. K3 reordered my bullets so the three most relevant experiences sit in the top third, rewrote phrasing to surface the JD's vocabulary where it honestly applied (I did run the pricing experiments; "pricing strategy" was never on my resume until it was true AND wanted), and killed anything irrelevant. One page, always. It did not add a single fact I didn't supply.
- The gap paragraph. For every hard gap it found, I wrote one honest sentence myself. K3's drafts were too defensive; a human acknowledging a gap plainly reads as seniority, and no model has landed that tone for me yet.
- Cover letter. Below.
- Interview prep. Below.
The Cover Letter Recipe
The first week, my K3 cover letters got zero replies, and rereading them I understood why: they were structurally perfect and completely interchangeable. So I built the opposite recipe. The letter must contain: one achievement with a number I can defend in an interview ("cut the onboarding drop-off from 34% to 19% over two quarters"), one specific connection to the company I couldn't fake ("I've used your API since the v2 beta and filed three bug reports"), and one sentence on why this role, this month. K3 drafts it, then we argue. My standing prompt ends with: "delete any sentence that could appear in a letter to a different company." The average letter went from 320 words to 140, and the callback rate went from 0% to 18% by week four.
Interview Prep From the Posting
This was K3's best trick and I haven't seen it written up elsewhere. Feed it the posting plus the company's recent changelog or blog index, and ask for: the five questions the hiring manager most likely wrote for themselves, and the two questions the panel will ask that the JD does not reveal (the gap edges). Then force it into roleplay — it interviewed me, I answered out loud, it critiqued structure, not content. By interview three, I had a written bank of 40 questions mapped to my own stories. Interview four and five, I wasn't rehearsing answers anymore, just recalling them.
The Numbers
| Metric | 2023 search (no pipeline) | This search (K3 pipeline) |
|---|---|---|
| Applications sent | 120 | 40 |
| Callbacks | 4 | 9 |
| First-round interviews | 4 | 6 |
| Final rounds | 1 | 3 |
| Offers | 0 | 2 |
| Token cost | — | $18.44 |
Five weeks from layoff to signed offer, start date in October. I accepted the smaller company — 40 people, product I actually use, $12K less than the other offer. The pipeline didn't make that decision; it just bought me the runway to be able to make it.
What I'd Do Differently
Two things. Start the tracking spreadsheet on day one — I lost a week re-reading postings to remember where I stood. And stop earlier on the tailoring depth: steps 1-3 get you 90% of the value for about 25 minutes per application. The applications where I polished for two hours beyond that did not measurably outperform. If you're in this situation right now, the other pieces I wrote that helped: the writing quality test that convinced me to trust it with prose, and the one-day SaaS build that became a portfolio talking point in every interview.
Frequently Asked Questions
Can Kimi K3 actually help you find a job?
It can compress the application work dramatically — my pipeline covered JD analysis, resume tailoring, cover letters, and interview prep for $18.44 in tokens. It cannot change your experience, and I drew hard lines against fabricating anything. The 6 interviews came from better targeting, not better fiction.
How do you make cover letters not sound AI-written?
My recipe: feed K3 two real things — a specific achievement with numbers from my career, and one concrete connection to the company (a product I use, a talk they gave). Then hold a short back-and-forth where I reject anything generic. A letter that could be sent to any company is deleted, not edited.
What is the best way to tailor a resume with AI?
Rearranging and re-emphasizing is fair game; inventing is not. I had K3 map the JD's requirements to my existing bullets, rewrite emphasis around what I actually did, and flag gaps honestly — then I handled the gaps in the cover letter myself instead of padding the resume.
Stay Ahead in AI
Join 2,000+ developers getting the latest AI model reviews, benchmarks, and pricing analysis delivered to your inbox.
No spam. Unsubscribe anytime.
