Chaos and Order

💡 왼쪽 원문을 읽으면서 오른쪽에 따라 써보세요. Tab 키로 힌트를 받을 수 있습니다.

원문 렌더가 준비되기 전까지 텍스트 가이드로 표시합니다.

Introduction — Why Talk About the Job Market Now

A recent Hacker News post titled "Please stop sending spam to job seekers" became a massive discussion thread. The moment a job seeker publishes a resume, they get buried under automated cold emails, fake AI-generated personalized recruiting pitches, and unresponsive ATS black holes — and hundreds of comments poured out the frustration. Similar discussions keep recurring on GeekNews as well. The core observation is simple: **with both sides of the hiring market armed with AI, the signal-to-noise ratio has never been worse.**

Here is the hiring landscape in the first half of 2026, summarized:

- A single job posting attracts hundreds or thousands of applications, because more candidates mass-generate resumes with AI and spray them everywhere.

- To cope with the flood, companies have made AI screening the default. By the time a human first looks at your resume, AI has already filtered the pool.

- With AI coding agents (Claude Code, Codex, Copilot and friends) now ubiquitous, "I can write code" is no longer a differentiator. Companies that allow AI tools during interviews are growing fast.

- As a result, paradoxically, **human-to-human trust — networking and referrals — has once again become the strongest channel.**

This article is a practical survival guide for that environment. Instead of vague advice, it focuses on formulas, scripts, and checklists you can copy and use immediately.

Understanding the Structure of the 2026 Hiring Market

Before building a strategy, look at the market structure. Here are the gates a single application must pass today:

[Applicant]

+------------------------+ Rejection rate approx. 70-90%

| 1. ATS keyword filter | (broken formatting, keyword mismatch)

+------------------------+

+------------------------+ Rejection rate approx. 50%

| 2. AI screening | (unclear impact, low role-fit score)

+------------------------+

+------------------------+ First time a human sees it

| 3. Recruiter review | (average dwell time 7-30 seconds)

+------------------------+

| 4. Hiring manager |

+------------------------+

Two facts follow from this:

1. **Stages 1 and 2 are read by machines.** You need machine-friendly structure and keywords.

2. **Referrals skip stages 1 and 2.** That is why the expected return on networking is far higher than it used to be.

This is also why Patrick McKenzie (patio11) and his classic essay "Do Not Call Yourself a Programmer" is even more relevant fifteen years later. He argued that engineers are "cost-saving or revenue-generating devices for businesses." In an era where AI writes the code, only people who can prove "I solved a business problem" — not "I write code" — survive.

Redesigning Your Resume — The Impact Bullet Formula

The Base Formula

Every experience bullet follows this formula:

[Action verb] + [what] + [how / at what scale] + [measurable outcome]

Formula: Action + Metric + Outcome

Before / After Examples

Compare weak bullets with strong ones:

[Before] Responsible for backend development of the payment service

[After] Redesigned the payment settlement batch around an async queue,

cutting processing time from 4 hours to 12 minutes and reducing

settlement-delay support tickets from 200 per month to zero

[Before] Contributed to improving code review culture

[After] Introduced PR templates and automated lint gates, reducing average

review round-trips from 3.2 to 1.4 and shortening deploy lead time

from weekly to 3 times per day

[Before] Used AI tools to improve productivity

[After] Built a Claude Code based migration agent pipeline that completed

the TypeScript conversion of 41 legacy APIs in 3 weeks

(manual estimate: 4 months)

[Before] Handled incident response and monitoring

[After] Redesigned p99 latency alert thresholds and wrote 12 runbooks,

reducing average overnight pages from 14 to 3 per month and

cutting MTTR by 38%

[Before] Developed internal search features

[After] Introduced embedding-based semantic search, lifting search CTR by 22%

and reducing the zero-result query rate from 18% to 6%

A Backup Formula When You Have No Numbers

A common objection: "My company never measured anything." In that case, hunt for substitute numbers in this order:

1. **Scale**: traffic, user counts, data volume, number of services ("a gateway handling 5 million requests per day")

2. **Comparison**: relative to before ("build time cut in half versus the old approach")

3. **Frequency**: how often a repetitive task happened ("automated a manual deploy done 5 times a week")

4. **Reach**: number of teams or systems affected ("a shared library used by 6 teams")

5. If you truly have nothing, **start measuring now.** Your last month before leaving determines the quality of your resume.

Differentiators in the AI Era — What to Add

There is a new set of signals recruiters and hiring managers look for in 2026 resumes.

1. Agent Operation Experience (Operating, Not Just Using)

"I use Copilot" is now a baseline, like having a driving license. Differentiated descriptions look like this:

- CLAUDE.md / AGENTS.md authorship: designed context documents and guardrails

so agents can work safely inside the team codebase

- Multi-agent pipelines: designed automated loops of code generation -> tests -> review

- Verification systems for agent output: built automated pre-merge validation gates

In current vocabulary, this is **loop engineering and context engineering** experience. The point is not "I write good prompts" but "I built a system where quality holds even when agents work autonomously for hours."

2. Evals Experience

Any team that has shipped an LLM feature has fought the evaluation problem. On a resume, write it like this:

- Built a 300-case evaluation dataset for an LLM summarization feature and

integrated regression evals into CI, blocking quality regressions from

prompt changes before deployment

3. System Design Ability

As code generation gets cheap, **the ability to decide what to build** appreciates in value. Make design docs, trade-off analyses, and ADR (Architecture Decision Record) experience explicit.

Differentiator Comparison Table

| --- | --- | --- | --- |

Getting Past ATS and AI Screening

Formatting Rules

ATS parsers are dumber than you think, and AI screeners are more conservative than you think.

- **Use a single-column layout.** Two-column layouts, text boxes, and experience written inside tables break parsing.

- **Standard section titles**: keep conventional headers like Experience, Education, Skills. Creative headers like "My Journey" get ignored by parsers.

- **Submit a PDF, but one with extractable text.** A PDF rendered as images is instant death. To check: try dragging and copying text inside the PDF.

- Do not put contact details in headers or footers. Some parsers skip them.

- Remove all icons, charts, and skill gauge bars.

Keyword Strategy

1. **Extract noun phrases** from the job description: tech stack, methodologies, domain terms.

2. Weave the keywords that intersect with your real experience **naturally into your experience bullets**. If they only appear in a skills section, AI screeners mark them as "no evidence" and dock you.

3. Use both the abbreviation and the full name **at least once**: "CI/CD (Continuous Integration)" style.

4. Tune the resume per posting. Not a full rewrite — adjusting **the top 3-line summary and about 10 keywords** is enough.

AI-Screening-Specific Tips

AI screeners go beyond keyword matching and infer "can this person solve the problems in the JD?" Therefore:

- Put a **3-line Professional Summary** at the very top, directly connecting your experience to the core problem of the role.

- Start each job with **the highest-impact achievement at that company** as the first bullet. Both AI and humans weight the first line.

- Never lie. Interviews in 2026 have evolved to dig deep into every line of the resume. An exaggeration that passes AI screening explodes in the interview.

Portfolio Strategy — GitHub and Demos

How to Organize GitHub

Recruiters spend an average of 2-3 minutes on your GitHub. You must deliver signal within that window.

1. **Curate your 6 pinned repositories.** A profile full of tutorial clones, empty repos, and forks is a net negative.

2. Each pinned README should communicate four things within 30 seconds:

- What problem the project solves (one sentence)

- One architecture sketch (even an ASCII diagram)

- How to run it (3 commands or fewer)

- Results or a demo (screenshot or video link)

3. **Commit history is part of the portfolio.** A history of meaningful commit messages builds more trust than a wall of "fix" and "update".

4. If code was AI-generated, do not hide it — write in the README **how you verified it**. In 2026, "taking responsibility for validating AI output" is a core evaluation criterion.

Demo Videos

A 90-second demo video is stronger than text.

Demo video structure (90 seconds)

0:00-0:10 Problem statement ("Deploys used to take 30 minutes...")

0:10-0:60 Core feature demo (real behavior, cut waiting time in editing)

0:60-0:80 One technical highlight only (do not try to explain everything)

0:80-0:90 Result summary + repository link

Link the video at the top of the README and in the projects section of your resume.

Networking — The Channel That Returned to the Throne

Why It Matters Again

In a market where applications pile up by the thousands, the hiring side also wants to finish before posting publicly. In practice, a large share of good positions are filled via referrals and personal recommendations **before the posting ever goes up**. Many postings are formalities with an internal favorite already chosen. This is the "hidden job market."

Networking does not mean handing out business cards at events. Practically, it is three things:

1. **Reactivating weak ties**: former colleagues, study group members, people you met in open source. It is a classic sociology finding that weak ties bring more opportunities than strong ones.

2. **Public work**: blogging, open source contributions, talks. An inbound channel that makes people come to you.

3. **Coffee chats with insiders at target companies**: gather inside information before applying, and convert to a referral when possible.

Referral Etiquette and Script

The most common mistake in referral requests is dumping the full weight of "vouch for me" onto the other person. A good request makes it **easy to say no**.

[Referral request script — to a former colleague]

Hey, long time no talk! Hope you are doing well.

I am currently looking for my next role, and the backend position at your

company (link) seems like a strong match for my background

(3 years in payment systems, large-scale batch optimization).

If it is not too much trouble, could I ask just two things?

1. Does the team and the actual work match what the posting describes?

2. If you think it is a fit, would you be open to referring me?

I am totally fine applying directly, so if you feel even slightly

unsure, please feel free to say no.

I will send over a polished resume in advance. Thanks so much for your time!

Key points:

- **Explicitly offer an exit** ("feel free to say no") — paradoxically, this raises acceptance rates.

- Provide **a one-line summary of your background** so the other person can easily write a recommendation reason.

- If the company pays referral bonuses, remember it benefits them too — no need to be overly deferential.

- Whatever the outcome, **always follow up and thank them.** A network is not single-use.

Interview Prep — What Changed in the AI-Tools-Allowed Era

The Transformation of Live Coding

The biggest change in 2026 interviews is the rise of **interviews that allow AI tools**. Memorized algorithm questions lost their discriminating power because AI answers them instantly, and these formats grew instead:

| Interview type | Then (2022) | Now (2026) |

| --- | --- | --- |

| Coding test | Whiteboard algorithms | Solving realistic tasks alongside an agent |

| Evaluation focus | Speed to correct answer | Problem decomposition, verification, tool orchestration |

| System design | Listing canonical components | Trade-off debate, cost estimation |

| Behavioral | Generic STAR questions | Resume deep-dives, stronger authenticity checks |

What evaluators watch in AI-allowed interviews:

1. **How you decompose the problem** — the person who throws the whole thing at the agent vs. the person who splits it into verifiable units

2. **How you verify the output** — do you write tests first, do you check edge cases yourself

3. **Behavior when stuck** — do you blame the tool, or form hypotheses and debug

How to practice: work with your agent while **narrating your reasoning out loud**. An interview is ultimately a broadcast of your thought process.

Behavioral Interview Prep

For every bullet on your resume, prepare this three-piece set:

Per bullet:

1. 30-second context version (why the work was needed)

2. 2-minute technical depth version (concretely what and how)

3. Failure / lesson version (one judgment you got wrong in the process)

Number 3 matters. With AI-inflated resumes flooding the market, interviewers use **the specificity of failure stories** as an authenticity signal.

Salary Negotiation Basics

Three Principles

1. **Never say a number first.** When asked for salary expectations, redirect with a question grounded in range research.

2. **The negotiation moment is after the offer.** Negotiating before an offer gives you no leverage.

3. **Competing offers are the best leverage.** Run your search in parallel whenever possible.

Scripts

[When asked for salary expectations]

"I would like to first understand the full compensation structure for the

role (base, bonus, equity). Could you share the band budgeted for this

position? If the role and responsibilities are right, I believe

compensation can be worked out within a reasonable range."

[When requesting an increase after receiving an offer]

"Thank you for the offer. I am very positive about both the team and the

role. That said, considering other processes I am currently in and market

data, I was expecting X in base salary.

If that can be adjusted, I am ready to sign right away. Would that be

possible to review?"

[When told an adjustment is difficult]

"Understood. In that case, beyond base salary, would options like a signing

bonus, an additional equity grant, or an early review at 6 months be

possible to consider?"

Key point: negotiation is not an adversarial act — keep the tone of **collaborating to make the terms work**. And the commitment signal of "I will sign if this is adjusted" helps your counterpart argue your case internally.

The Job Search Kanban — Running a Tracking System

Once you pass 10 applications, your head cannot manage it. Run a kanban.

+-----------+-----------+-----------+-----------+-----------+-----------+

+-----------+-----------+-----------+-----------+-----------+-----------+

+-----------+-----------+-----------+-----------+-----------+-----------+

What to record on each card:

- Posting link, application date, resume version used

- Recruiter name and contact

- Next action and deadline (e.g., follow-up email by 6/20)

- Post-interview retro notes (questions, what went well, what did not)

Operating rules:

- **Fix a weekly cleanup slot** (e.g., 30 minutes Friday afternoon).

- Cards with no response for 2 weeks get one follow-up email, then move to Closed. Lingering hope drains energy.

- Interview retro notes are the best textbook for your next interview. Write them the same day, always.

Managing Your Mental State

A job search is a months-long marathon, and rejection is the default outcome of the game.

- **Measure only what you control.** Target behavioral metrics like "5 tailored applications and 2 coffee chats this week," not "number of offers."

- **Rejection is data.** If you keep failing at the same stage, improve only that stage. Many resume rejections: fix the resume. Many first-round rejections: practice technical explanations.

- **Do not job-hunt all day.** Splitting the day — 3 hours of job search in the morning, learning or side projects in the afternoon — decouples your self-worth from search outcomes.

- Ghosting with no rejection notice is now routine in this market. Remember it is system overload, not a judgment of you as a person.

Etiquette for the Hiring Side — Stop the Spam Cold Emails

For balance, the hiring side deserves a section. The gist of the HN post mentioned at the start: the moment job seekers publish their resume, they get flooded with fake AI-personalized recruiting emails. Messages open with "I was deeply impressed by your experience with X" but pitch positions unrelated to their background, and replies get answered by bots. Job seekers are exhausted enough already.

A minimum etiquette for recruiters and sourcing engineers:

1. **If you did not read it, do not pretend you did.** Recipients instantly detect AI-generated "personalization." An honest template is better.

2. **Disclose the full position details upfront.** Company name, salary band, tech stack, work arrangement. "Let me explain on a call" holds the other person's time hostage.

3. **Do not re-send after a rejection reply.** Auto-sending a 7-email sequence to the same person is not marketing — it is spam.

4. **Always send rejection notices.** Companies that ghost candidates after full interviews get their reputation immortalized in developer communities.

Job seekers benefit from knowing this dynamic too: the quality of cold emails you receive is a first-pass filter on the company, and a genuinely personalized contact (one that concretely mentions your blog post or open source contribution) is a rare signal worth a polite reply.

Pitfalls and Critical Perspectives

For balance, here are the common counterarguments:

- **"Is keyword optimization not just an arms race?"** Yes. When everyone optimizes, only the average rises. That is why this article treats ATS tactics as a necessary condition and networking plus public work as the sufficient condition. Differentiation comes from the latter.

- **"In the AI era, does the resume itself not become meaningless?"** Long term, evaluation will likely shift from resumes toward verifiable work and realistic collaboration simulations. But as of 2026, most company processes still start with a resume. In the transition period, you need both.

- **"Networking is unfair to introverts"** is a fair criticism. That is why this article emphasizes **asynchronous networking** channels like writing and open source. They work without event-floor sociability.

- **Some markets allow no negotiation at all.** At large companies with rigid bands or batch-hiring organizations, there is little room. Scripts are not magic — they are tools for when you have leverage.

Final Checklist

Resume

- [ ] Does every experience bullet follow the action + metric + outcome formula

- [ ] Is the first bullet of each job the highest-impact one

- [ ] Single-column layout, PDF with extractable text

- [ ] Does the top 3-line summary connect directly to the target role

- [ ] Are JD keywords woven into bullets with evidence

- [ ] Do agent operation, evals, and system design experience show

- [ ] Zero lies or exaggerations (can it survive an interview deep-dive)

Portfolio

- [ ] Are your 6 GitHub pins curated

- [ ] Does each README convey problem-structure-run-result within 30 seconds

- [ ] Does your flagship project have a 90-second demo video

Process

- [ ] Are you running a job search kanban with a weekly cleanup

- [ ] Do you have weekly behavioral targets (N applications, N coffee chats)

- [ ] Do you write interview retros the same day

- [ ] Did you offer an exit when requesting referrals

- [ ] Did you avoid naming a number before the offer

Closing

The 2026 hiring market is undeniably full of noise. But the louder the noise, the more valuable real signal becomes. Verifiable impact, public work, and human trust — these are the three things AI cannot mass-produce. Fixing one resume bullet and requesting one coffee chat are both part of accumulating that signal. I hope the checklists in this article serve as a map for that journey.

References

- Hacker News discussion: Please stop sending spam to job seekers — https://news.ycombinator.com/item?id=48370330

- Patrick McKenzie, Do Not Call Yourself a Programmer — https://www.kalzumeus.com/2011/10/28/dont-call-yourself-a-programmer/

- Hacker News: Who is hiring? monthly threads — https://news.ycombinator.com/submitted?id=whoishiring

- GeekNews (developer hiring and career discussions) — https://news.hada.io/

- The Pragmatic Engineer Blog (hiring market analysis) — https://blog.pragmaticengineer.com/

- levels.fyi (compensation data) — https://www.levels.fyi/

- Julia Evans, Questions to ask your interviewer — https://jvns.ca/blog/2013/12/30/questions-im-asking-in-interviews/

- patio11, Salary Negotiation guide — https://www.kalzumeus.com/2012/01/23/salary-negotiation/

- GitHub Docs: Pinning items to your profile — https://docs.github.com/en/account-and-profile/setting-up-and-managing-your-github-profile/customizing-your-profile/pinning-items-to-your-profile

- Anthropic: Claude Code official docs — https://docs.anthropic.com/en/docs/claude-code/overview