What the Recruiter Can't Do
Frederick P. Brooks Jr., No Silver Bullet (1986):
There is no single development, in either technology or management technique, which by itself promises even one order-of-magnitude improvement within a decade in productivity, in reliability, in simplicity.
Brooks was writing about software. The same humility belongs on a home-built recruiter. This system will never make a phone call, read a room, or know that the hiring manager just had a rough quarter. It filters, drafts, and tracks. It does not replace the parts of a job search that are actually about people.
This chapter is the honest limits section: what to keep doing yourself, and why treating the machine as an amplifier instead of a replacement is what keeps it useful instead of creepy.
What the system does
Section titled “What the system does”Full stop. If a task is not on this list, do not pretend the overnight run handled it.
| Capability | What “done” looks like |
|---|---|
| Filter | Dealbreakers applied, survivors scored, tiers assigned (chapters 2–5) |
| Report | Dual artifact ready for coffee: human skim + machine fields (chapter 6) |
| Draft | Outreach and resume variants that clear an edit bar, not a rewrite bar (chapters 7–8) |
| Remember | Status, outcomes, and reason codes available to the next sweep (chapter 9) |
| Signal | Thin streaks and pattern counts visible enough to force a strategic review (chapter 10) |
That is a lot. It is also bounded. The product is protected attention and continuity, not a hired human who knows your industry’s gossip.
What the system is good at, in one line: repeating judgment you already articulated, at a cadence you already chose, without getting tired of your own rules.
What it can’t do
Section titled “What it can’t do”Be explicit, or the fantasy creeps back in every quiet week.
It cannot read a room. Tone on a call, political weather inside a company, whether the founder is hiring from conviction or from board panic. Your notes field can store what you learned after. It cannot sense it before.
It cannot make the call (or the DM that needs courage). It can put a draft in front of you. Hitting send is still a human act. So is hearing “no” without rewriting the rubric in anger.
It cannot know timing you never encoded. Hiring freezes, budget cycles, “we just hired someone,” a rough quarter, a stealth pivot. Public signals help sometimes. Most of the useful timing is relational.
It cannot build real rapport. Peer-to-peer voice samples make drafts sound like you. Relationship is what happens across multiple human touches: listening, follow-through, reputation in a network. Automating that past a point becomes spam with better grammar.
It cannot negotiate for you. Comp, scope, side-role tolerance, start date, equity story. The dealbreaker list tells you what to ask. You still ask. You still decide.
It cannot invent your values. If you never wrote the culture criteria, the score will launder whatever the JD implies. Garbage criteria in, confident garbage out.
It cannot make a bad company good. A high score means fit to your written rules, not moral clearance, not “this will work out.”
It cannot replace grief, ego, or rest. Searches happen around layoffs, identity hits, and family constraints. A green morning report does not metabolize that. Expecting it to is how people build ever-more-elaborate automation instead of taking a walk.
Chapter 1’s bulk-apply tools fail because they deny these limits. They sell replacement. This book sells amplification. The minute you blur the two, you recreate the thing you left the job boards to escape: volume without judgment, now with a pipeline diagram.
Amplifier, not replacement
Section titled “Amplifier, not replacement”The framing that keeps the system useful:
Amplifier: You articulate dealbreakers, criteria, voice, and cadence. The machine applies them while you sleep, drafts while you focus, remembers while you move on. You spend human time on human leverage.
Replacement: You outsource the search to agents and hope intent emerges from keywords. You send what you would not say. You optimize for count of applications. You hide from calls behind dashboards.
| Move | Amplifier | Replacement (creep / waste) |
|---|---|---|
| Outreach | Draft → edit → send as yourself | Auto-send 40 “personalized” DMs |
| Resumes | Base + fifteen-minute variant | Bulk rewriter invents a new career nightly |
| Follow-up | One bump logged in memory | Sequence that stalks a ghost |
| Scoring | Rubric you could defend out loud | Model vibes a fit score with no weights |
| Quiet market | Diagnose and recalibrate | Scrape harder, lower the floor in secret |
Creep is not only ethical. It is practical. People can smell industrial outreach. So can you, when you reread a sent folder full of messages you do not remember approving. The edit bar in chapter 7 is an ethical control as much as a quality control: if you would not say it, it does not ship.
Amplifier also means the system is allowed to be incomplete. Missing a feature is fine. Missing a judgment you refused to write down is not something an agent should paper over.
What to keep doing yourself
Section titled “What to keep doing yourself”Put your time where the needle moves.
1. The clarity interview and the dealbreaker list. Only you know the floor you will not drop (chapters 2–3). Revisit after thin streaks; do not delegate the revisit to a model that wants to be helpful.
2. The Strong Yes send. Edit the draft. Hit send. Own the first impression.
3. The live conversation. Phone, video, coffee. Ask the gated questions early (comp, remote, side role, hands-on scope). Listen for what the rubric cannot hear.
4. Relationship maintenance. Thank-yous, intros, updates to people who helped, honest withdrawals when you take another path. Memory can remind you; it cannot mean it.
5. Negotiation and decision. Offer math, trade-offs, yes/no. The report does not sign.
6. Ethics and disclosure. Active companies, conflicts, what you will not pretend. If the base resume hides something your dealbreakers require you to say on call one, that is on you (chapter 8).
7. Rest and recalibration. Chapter 10’s strategic review is a human meeting with yourself (and maybe one trusted peer). Schedule it. Do not skip it because the sweep “might pick up tomorrow.”
A daily split that stays sane
Section titled “A daily split that stays sane”| Block | Owner | Job |
|---|---|---|
| Overnight | System | Sweep, gate, score, report |
| Coffee (15–20 min) | You | Tier decisions, no raw-board browsing |
| Outreach block | You + drafts | Edit/send Strong Yes; queue Yes |
| Weekly review | You + memory | Patterns, one change, thin-streak check |
| Calls | You | Everything the machine cannot sense |
If your calendar inverts (hours of prompt-tweaking, no calls, no sends), the amplifier has become a hobby. Pause feature work. Ship the next human touch.
The limits are the point. A recruiter that claimed to do everything would be lying in the same genre as the job boards that claim abundance is the same as fit. Hold the line, and the system stays a tool you trust. Chapter 12 is how to build the smallest version that still deserves that trust.
Next: Fork It →