Claude can describe a good hire. It cannot find one.

    A general model is excellent at language and reasoning. Put it in your Slack and it will write a job description in nine seconds. Ask it who to hire and it has nothing to look at, because it has no index of people, no scoring model and no view of your pipeline.

    Start for freeBook a demo

    $100 of credits on us. No card required.

    The short answer

    Putting a general model in Slack gives you a very capable colleague who has never recruited, has no candidate database and forgets your pipeline the moment the thread scrolls away.

    Candidate dataMy own index of the open web, plus every candidate already in your ATS
    RankingDeterministic score with the evidence attached to each line, stable across runs
    Recruiting knowledgeBuilt in: scope, comp bands, seniority signals, disqualifiers and skill inference ship with me

    What you get either way

    Both sides, written plainly. The one on top is mine.

    By choosing me, you get

    A recruiter who arrives already trained

    An index, not an opinion

    • My own crawl of repositories, portfolios, papers and company writing
    • Your ATS re-read on every brief, so past applicants resurface
    • Names that never wrote the keyword you searched for

    Judgement that ships with me

    • Scope, comp band, seniority signals and disqualifiers derived from one sentence
    • Change one dimension and I recalibrate the whole brief
    • No prompt library to write or maintain

    The work after the list

    • Drafted outreach and follow ups, sent only after you approve
    • Scheduling across calendars, in the channel
    • ATS records kept current without anyone typing them
    #hiring-backend · shortlist
    BriefSenior payments engineer, Berlin or remote EU
    Searched1,400 parallel queries across my index plus your ATS
    Returned23 scored candidates, evidence linked per line
    Next8 drafts waiting for your approval
    By choosing a general model, you get

    A brilliant generalist with no candidates

    Nothing to search

    • No people index of its own
    • Answers are only as good as what you paste in
    • Connected tools return their own limited snapshot

    Judgement you supply

    • Hiring standards live in your prompt, not in the tool
    • Every recruiter on the team writes a slightly different one
    • Rankings can shift between runs of the same question

    Work that stays yours

    • Scheduling, chasing and notes are still manual
    • The ATS gets updated by a human, or not at all
    • Copy and paste between the thread and the tools
    Thread · general assistant
    BriefSenior payments engineer, Berlin or remote EU
    ReturnedA search strategy and a boolean string
    CandidatesNone it can verify
    NextYou open another tab

    The difference in one table

    ClaudeLaidback
    Candidate dataNo people index. It reasons over what you paste in or what a connected tool returnsMy own index of the open web, plus every candidate already in your ATS
    RankingFluent judgement that can change between two runs of the same promptDeterministic score with the evidence attached to each line, stable across runs
    Recruiting knowledgeGeneral reasoning. The hiring judgement has to come from your promptBuilt in: scope, comp bands, seniority signals, disqualifiers and skill inference ship with me
    SetupYou write the prompts, wire the tools and maintain them as roles changeInvite me to the channel and describe the role in one sentence
    Follow throughDrafts text. Scheduling, follow ups and ATS hygiene stay with youDrafted outreach, follow ups, scheduling and ATS write back, with you approving
    Where the work happensA chat thread, then a lot of copy and pasteSlack, Teams or WhatsApp, on top of 1,000+ tools including your ATS and calendar

    What a general model cannot do for a req

    It has no one to search

    A model without an index can only reason about people you already know. Most of my shortlist has no searchable profile anywhere, because I build profiles from shipped work rather than self description.

    Its answers move

    Ask the same question twice and a general model can rank differently. A hiring manager will notice. My score is derived the same way every time, with the reasoning written next to the name.

    Nothing gets finished

    The half of recruiting that eats the week is scheduling, chasing, notes and keeping the ATS honest. That work needs write access and a memory of the pipeline, not a better paragraph.

    Where Claude still wins

    Anything outside hiring, and anything where you want an open ended thinking partner. It is a better general writer and a better generalist. Keep it. We are not competing for the same job.

    #hiring-backend
    Marcin
    Marcin

    We already have an AI assistant in Slack. Why add you?

    Laidback

    Ask it for eight senior payments engineers in Berlin with evidence. It will give you eight plausible names or a search strategy. Neither is a shortlist.

    Laidback

    I run the brief across my own index of the open web, score every candidate the same way twice, and write the reasoning next to each name.

    Index and scoring, not prose
    Marcin
    Marcin

    So keep both?

    Laidback

    Yes. Keep it for everything that is not hiring. Hand me the reqs.

    What people ask when they already run an AI assistant

    Can we keep Claude and add you?

    Yes, and most teams do. It handles thinking, writing and everything outside hiring. I handle the reqs. Different jobs, same Slack.

    Could we just connect Claude to a sourcing API?

    You can, and some teams have. You end up owning the index choice, the scoring logic, the recalibration and the maintenance. That is the product I built, not a weekend integration.

    Is your model better than Claude?

    Wrong question. The models are commodity. What is not commodity is my index of the open web and a scoring model that returns the same answer twice.

    Where does your candidate data come from?

    My own index of the open web: repositories, portfolios, publications, personal sites, company writing and talks, plus the candidates already sitting in your ATS. Profiles are built from what people ship, not from a licensed snapshot every vendor buys.

    Do you send anything on your own?

    No. I draft, you review, you send. Where a habit runs on its own you pick the mode: Autopilot or Ask approval.

    I am the opposite trade. Narrow on purpose: one job, with the index, the scoring and the recruiting judgement built in, so nothing has to be prompted into existence first.

    On the independent people search benchmark I score 97.80 out of 100. The method and every scored query are published on the benchmark page.

    One job, learned properly

    I do recruiting and nothing else. The bar, the etiquette, the order of the stages and the follow up nobody remembers are all built in rather than configured.

    Built for hiring, not adapted to it

    General agents learn recruiting from your prompts. I was built on it. The bar, the stages, the etiquette of a first message are already in me.

    Skills that come from real calibration

    I learn your bar from the people you said yes and no to, then score against it the same way twice. Judgement you can inspect, not a similarity number.

    Workflows that finish the job

    Sourcing, evidence, drafts, follow up, scheduling, ATS write back. The parts nobody enjoys sit in the same run, so nothing stops at a list of names.

    Context that keeps compounding

    Every brief, rejection and debrief stays with me. Role three starts from what roles one and two taught me.

    Where to go next

    I run on top of your ATS, I do not replace it

    Templates you can use today

    Laidback

    Give your assistant the writing. Give me the reqs.

    Start for freeBook a demo

    One open role, one sentence. Compare what comes back with what your assistant returns.