Ask HN: Can LLMs change the hiring economics of legacy engineers?
SubscribeSign in
Ask HN: Can LLMs change the hiring economics of legacy engineers?<br>A small anonymized field experiment about screening, legacy engineers, and LLM-enabled adaptation.
Hryhorii Kuzyk<br>Jun 19, 2026
Share
I am running a small side experiment about hiring, legacy experience, and LLMs as a possible candidate multiplier.<br>This is not a study and not a statistical claim. It is one real anonymized case that I want to test to the end.<br>For the last three months, I have been testing a simple hypothesis: can a candidate with a “dead” or poorly matched tech stack, who actively uses AI tools, make it through a standard hiring funnel at a real software company?<br>Stage 1: CV screening
I submitted a real person’s CV, with name and identifying details changed, for two middle-level roles: Node.js and Vue.js.<br>The candidate’s actual background: enterprise / legacy systems, architecture, integrations, Salesforce-like ecosystem, SAP business analysis, and complex corporate workflows. Adjacent skills, but not the stack the job postings asked for.<br>Two recruiters independently reviewed the CV for both roles.<br>Result: No Pass for both.<br>The reasoning was specific and accurate: no MongoDB, no clearly confirmed hands-on Vue.js skill, Node.js only in a supporting role inside integration workflows rather than as full backend development.<br>So at the CV level, the rejection was reasonable.<br>Stage 2: Technical interview
The case then moved from “does the CV match the vacancy?” to “what can this person actually do as an engineer?”<br>It took over a month to set up. The technical interview was not narrowly tied to one exact stack. It focused more broadly on backend, frontend, fundamentals, architecture, and engineering thinking.<br>Result: positive technical feedback.<br>The interviewer described the candidate as strong on the backend side, with solid architectural understanding. There were gaps in stack breadth, partly due to years in legacy / enterprise environments. But the engineering fundamentals were strong, and the interviewer’s assumption was that the candidate would likely pick up new technology quickly.<br>At CV screening level: No Pass.<br>At technical interview level: positive signal.<br>CV screening measures role fit; a technical interview can reveal adaptation potential.<br>Why this matters now
What gets published about “AI performance” often does not show the real picture.<br>Roughly, I see three extremes.<br>1. People downplay it. They get real gains from AI, but do not surface them, either because they do not want their work to look “too easy” or because they fear their role will look less essential.<br>2. People flood managers with unverified technical output. Slop dressed up as productivity. Volume mistaken for quality.<br>3. People simply ignore AI, as if the tool does not exist.
None of these show what real performance looks like when someone uses LLMs competently.<br>Final stage
The question is no longer just whether this candidate can do the job. That is too narrow.<br>The more interesting question is whether this person can become an internal LLM multiplier.<br>Not an AI evangelist.<br>Not someone selling “AI transformation” slides.<br>But a practical specialist with architectural thinking who understands complex systems, sees where a team loses time, and can actually raise team performance through better task formulation, review, documentation, onboarding, routine reduction, and better use of internal knowledge.<br>For the final stage, I am hiring the candidate on a fixed-price contract below market rate.<br>The test is simple: give them a real slice of work, not an abstract take-home, then measure whether there is a real improvement in quality, speed, or cost.<br>Main question:<br>How long, and at what price, does a risky candidate with the wrong stack become someone who genuinely raises a team’s performance?<br>Not through motivational talks. Not through AI magic. Through actual work.<br>Field test
I would also be interested in testing this with a real team of 10–20 people.<br>No upfront investment.
Share
Discussion about this post<br>CommentsRestacks
TopLatest
No posts
Ready for more?
Subscribe
© 2026 Hryhorii Kuzyk · Privacy ∙ Terms ∙ Collection notice<br>Start your SubstackGet the app<br>Substack is the home for great culture
This site requires JavaScript to run correctly. Please turn on JavaScript or unblock scripts