Refolk
November 7, 2026·10 min read

Hiring Assistant 2 Can't See GitHub. Source the Rust Engineers It Misses

LinkedIn Hiring Assistant 2 only ranks LinkedIn members and your ATS. Here's how to source the Rust, ML infra and GPU engineers it can't see.

LinkedIn Hiring Assistant 2Hiring Assistant 2 limitationsLinkedIn Recruiter AI agentsourcing candidates outside LinkedInAI sourcing tools for recruiters
Hiring Assistant 2 Can't See GitHub. Source the Rust Engineers It Misses

If you're hiring for Rust, ML infra or GPU roles, Hiring Assistant 2 will build your shortlist from the same place every other LinkedIn Recruiter customer's shortlist comes from: LinkedIn members and your own ATS. It can't see an engineer whose Rust lives in GitHub commits, crates or Hacker News posts rather than on their profile, and those are the skills where this month's HN hiring demand is concentrated.

The upgrade starts showing up in Recruiter Corporate and RPS+ accounts this November and switches on automatically for English-language customers at no added cost. It remembers how you hire, scores fit and interest, and runs voice pre-screens. None of that widens the pool.

What does LinkedIn Hiring Assistant 2 actually add?

Hiring Assistant 2 ranks the same pool LinkedIn always searched, just faster and with more personal context. LinkedIn announced it on September 29, 2026, in a post from Dan Reid (VP Product) and Prashanthi Padmanabhan (VP Engineering). Reid's selling point at Talent Connect 2026 was that customers don't need to migrate anything.

The LinkedIn Recruiter AI agent is an add-on inside LinkedIn Recruiter that ranks candidates and sends outreach for the recruiter. Version 2 adds:

  • Recruiter memory, so the agent keeps your preferences from one project to the next.
  • Fit and candidate-interest signals on each recommendation.
  • Verification inside partner ATSs.
  • Voice pre-screening of candidates.
  • Talent-pool management across Greenhouse, Lever and SmartRecruiters.

LinkedIn's own figures for early v1 users were more than 4 hours saved per role and 62% fewer profiles reviewed. Both numbers measure review effort. Neither one says the agent found anyone you couldn't have found already. That difference is what the rest of this piece is about.

Hiring Assistant 2 limitations: what can't it see?

Hiring Assistant 2 only recommends LinkedIn members and people already in your applicant tracking system. It doesn't search GitHub, Hacker News, personal sites or the rest of the open web. HeroHunt sells a competing product, so it's fair to discount its framing. Still, its description ("still finds only LinkedIn members and people already in your own applicant tracking system") matches an independent guide on theprofessions.cloud, which says the agent reaches "LinkedIn members and your ATS, not the open web."

The other limits are worth knowing before you plan a quarter around the tool:

  • Contract gate: only available on Recruiter Corporate or RPS+ contracts that carry the add-on.
  • ATS coverage: Greenhouse, Lever and SmartRecruiters now. iCIMS and Workday, the big-enterprise systems, are still listed as "coming."
  • EU projects: instant outreach is turned off for projects located in the EU.
  • Outreach mechanics: batches of up to 25 candidates, up to 35 qualifications per project, and every message draws on InMail credits and Hiring Assistant credit limits.

The ATS integration is the most misunderstood part of the release. A talent pool built from your ATS brings back people you've already met. That's useful for rediscovery, since a silver-medalist from last spring might be ready now. But it adds nobody new to the market you can reach. Your ATS holds your past, and the agent reads that past more efficiently.

Why are the hottest skills the hardest for the agent to find?

The agent ranks what people write on their profiles, and the scarcest skills are the ones engineers are least likely to write down. A skill tag is a self-declared keyword on a profile. A ranking agent can only match a skill someone has declared or that shows up in their profile text.

In Refolk's index of professional profiles, the US senior-level counts by skill tag look like this:

Skill (US, senior)Profiles in Refolk's indexvs. Python
Python183,8951x
CUDA5,1362.8% (about 1 in 36)
Rust1,3260.72% (about 1 in 139)
1,326
US senior-level profiles tagged Rust in Refolk's index
Against 183,895 tagged Python, that is about one Rust tag for every 139 Python tags.

These counts show how often people tag a skill. They don't show how many people actually have it, and that gap is the point. Picture a senior systems engineer whose profile says "Staff Engineer" and little else. If their Rust lives in commit history, published crates and HN comment threads rather than a skills section, the agent has nothing to match. It can't rank evidence it can't read.

Don't read this as "Rust engineers aren't on LinkedIn." Many may well be. The problem is under-tagging: the profile exists but doesn't say what you're searching for. And the mismatch is sharpest where demand is hottest. Rust is one of the top demand clusters in November's HN hiring posts, yet even CUDA, itself scarce at 5,136 senior tags, outnumbers it 3.9 to 1 in the same index.

Same pool, same model, same shortlist

When thousands of companies run similar agents over one member graph, their top picks converge, and the scarcest candidates get flooded. More than 20,000 companies use LinkedIn's agentic hiring products, including AMD, Cisco, Palo Alto Networks and Zillow. That figure covers all of LinkedIn's agentic tools, not just Hiring Assistant seats, but the mechanism is the same either way: one graph, comparable ranking logic.

Recruiter memory doesn't change this. Memory changes how the agent weights candidates for you. It doesn't change who the agent can see. If you and a competitor both want a senior Rust engineer in the US, you're both ranking from the same 1,326-tag neighborhood, and your shortlists will overlap heavily near the top. Ruzora's review points out that the most-wanted engineers already "get the same InMail from dozens of companies."

Refolk's small samples show who sits at the top of these pools. Senior Rust-tagged US profiles include people at Cloudflare, Figure, SpaceX, OpenAI and F5. In the senior CUDA sample, NVIDIA is the most common employer, followed by Meta, Databricks and Amazon. The top titles in both are Staff, Principal and "Member of Technical Staff." These come from 25-profile samples, so treat them as illustrative. They still describe the people every agent will surface first: well-placed, senior, and already fielding the same message from other companies.

An agent that ranks the same pool for everyone doesn't give you an edge. It gives you a seat in the same queue.

This is how a hard role turns into a long one. The agent hands you a confident top 25, replies are thin because those 25 may have heard the same pitch this week, and you rerun the project with looser qualifications. After a month you've used the credits, the hiring manager is talking about an agency fee, and the role is no closer to filled.

Does automation raise your outreach ceiling?

Hiring Assistant 2 doesn't raise your outreach ceiling, because agent messages draw on the same InMail credits as manual ones. It speeds up choosing who to contact without increasing how many people you can contact. Recruiter Corporate includes up to 150 InMails per seat per month, and the agent sends in batches of up to 25.

ConstraintFigure
InMails per Recruiter Corporate seatUp to 150 per month
Agent outreach batch sizeUp to 25 candidates
Qualifications per projectUp to 35
UK add-on licence (promo ended June 30, 2026)£1,575 to £2,079 per year
UK Recruiter seat£6,350 to £8,925

US pricing isn't published. The UK rate card is enough to show what's going on: you pay per seat for a fixed messaging budget, and the agent spends it faster. On scarce-skill searches, faster spending on a contested shortlist is a poor trade.

Email, a reply under someone's HN post, or a message that references a specific GitHub commit doesn't use any credits. Those channels are also where the under-tagged engineers from the earlier section are easiest to identify.

Where to source candidates outside LinkedIn

For systems, Rust and GPU roles, start where the work is visible in public: Hacker News hiring threads, GitHub and the open web. Sourcing candidates outside LinkedIn means finding people through evidence of their work instead of their self-description.

November 2026 HN "Who is hiring?" posts clustered around:

  • Systems engineering
  • ML infrastructure
  • Rust
  • Python and PyTorch
  • GPU clusters
  • Kubernetes

This kind of demand often shows up on HN before it shows up on LinkedIn. Thunder Compute (YC S24) posted for C++ systems engineers to build a GPU "hypervisor." Another November post was for an infra role running multi-cluster GPU training on Slurm, with "a compute budget of tens of millions a year."

The supply side is busy too. The October 2026 "Who wants to be hired?" thread drew 609 comments. The November pair went up on November 2 because the 1st fell on a Sunday. The off-LinkedIn channels worth working:

  1. HN "Who wants to be hired?" threads, plus third-party search tools such as hnwork.app and gethnjobs.com (a fuller list is at nchelluri.github.io/hnjobs).
  2. GitHub: contributors to the repos your stack depends on, crate and package maintainers, and people with recent commits to GPU scheduling or Slurm tooling.
  3. The open web: conference talks, technical blogs and personal sites where engineers explain what they actually built.

Doing this by hand is slow. You're joining a GitHub handle to a name, a name to an employer, and an employer to a current role, then checking the person isn't already in your ATS. That stitching across sources is the gap Refolk is built to close. You describe the person in plain English and get back a ranked list drawn from GitHub, LinkedIn and the open web, including people whose profiles never mention the skill.

The useful part is the overlap check. If a candidate shows up in Refolk and also ranks near the top of your Hiring Assistant project, assume they're contested and write accordingly. If they show up only in Refolk, you may be the first recruiter who has read their code.

A sourcing plan built around the blind spot

Use Hiring Assistant 2 for the well-tagged pool and spend your own time on the pool it can't reach. The two approaches work together, and the split should follow how scarce the skill is.

  1. Let the agent handle commodity skills. For roles anchored on Python, with 183,895 senior US tags in Refolk's index, the agent's ranking and the 62% review reduction LinkedIn reports are real gains.
  2. Sweep outside LinkedIn first for scarce skills. For Rust (1,326 tags) and CUDA (5,136), build the external list before you open a Hiring Assistant project, so the agent's top 25 doesn't anchor your thinking.
  3. Treat overlap as a signal. Anyone who appears in both your external list and the agent's shortlist is being pitched by other agents too. Lead with something they can't get elsewhere.
  4. Spend InMails on the contested few. Use free channels (email, GitHub-referenced notes, HN replies) for everyone else and keep the 150 monthly credits for people only reachable on LinkedIn.
  5. Weight external sourcing higher for EU projects. With instant outreach off there, a European team gets less from v2 than a US team, so the off-LinkedIn work pays back more.
  6. Track replies by source. After one quarter you'll know whether the agent's shortlist or your external sweep produces the hires for each role family.

Agent settings won't separate recruiters when every eligible LinkedIn customer gets the same agent at no added cost. Where they search will.

FAQ

Does LinkedIn Hiring Assistant 2 cost extra?

LinkedIn is rolling it out automatically and at no added cost to existing English-language customers from November 2026. Access is still limited to Recruiter Corporate or RPS+ contracts that carry the Hiring Assistant add-on. The public UK rate card listed the add-on at £1,575 to £2,079 per licence per year under a promotion that ended June 30, 2026. US pricing isn't published.

Can Hiring Assistant 2 search GitHub or Hacker News?

No. It recommends only LinkedIn members and people already in your ATS. If an engineer's Rust or CUDA work is visible only in commits, packages or HN posts, the agent won't surface them unless their LinkedIn profile or ATS record also reflects it.

How does Hiring Assistant 2 work for EU-based roles?

Instant outreach is turned off for projects located in the EU, so European teams get less automation than US teams from the same release. That makes off-LinkedIn sourcing relatively more valuable for EU hiring, especially for scarce systems and GPU skills.

Should teams drop Hiring Assistant 2 for other AI sourcing tools?

Usually not. It's free to eligible customers, and LinkedIn's v1 figures show real cuts in review time, which matter most on well-tagged, high-volume skills. The better approach is to pair it with a tool that reaches GitHub and the open web, and to save its InMail-limited outreach for candidates who can only be reached on LinkedIn.

The data behind this

Try it on the search you came here for

Stop building boolean strings. Just describe the person.

Type one sentence. I plan the search, read GitHub, public LinkedIn and Crunchbase records, and the open web as it is right now, and hand back a ranked list with the reason next to every name.

  1. 01Describe them

    One plain sentence. Role, city, stack, stage, whatever matters to you.

  2. 02I read the web live

    GitHub, public LinkedIn and Crunchbase records, the open web. Not a database that went stale last quarter.

  3. 03You read the shortlist

    Ranked, with the reasoning under every name. Open a profile, ask a follow-up, narrow it down.

  • No boolean, no filters, no seat to buy. One box.
  • Read at search time, so a profile updated yesterday counts today.
  • Every step visible as it runs, every name with its reason.

500 free credits on sign-up. No card, no demo call. See real searches.

Read next