LinkedIn Hiring Assistant 2 Can't See GitHub. That's 99.81% of Rust.

On September 29, 2026, LinkedIn used the Talent Connect NYC stage to announce Hiring Assistant 2, an upgraded AI agent with memory, "do what I meant" reasoning, and new reach into screening and interview coordination. English-speaking Hiring Assistant customers get the upgrade free in November. The pitch is good. The ceiling has not moved an inch.

What LinkedIn actually shipped at Talent Connect

Hiring Assistant 2 is a smarter in-network agent, not a wider one. LinkedIn VP of product management Dan Reid framed the headline feature as intent parsing: the agent should "not just do what I told it to do. It does what I meant to tell it. That's a big difference. Do what I meant, not what I said."

The concrete deliverables, from the keynote and the day-of coverage:

• Memory and personalization. Each requisition, candidate, and hire teaches the agent recruiter preferences and habits.
• Expanded agentic workflows. HA2 now reaches into manager feedback, candidate screening, and interview coordination, not just sourcing.
• Free auto-upgrade in November. Every English-speaking Hiring Assistant customer wakes up on the new version with no procurement cycle.
• Macro framing. LinkedIn cited its own figure that AI created almost 2 million new jobs in three years, that global hiring is down 30% while applications per applicant are up 30% versus pre-pandemic, and that 64% of US TA pros say it is getting harder to know who to trust when assessing candidates.

Talent Connect Summit itself is small, roughly 1,200 invite-only senior TA leaders. That is a tight room with high word-of-mouth velocity, so adoption of the agentic framing will move faster than the actual feature rollout.

The one chart that explains the whole announcement

LinkedIn's reasoning upgrade matters most for roles where the pool is dense on LinkedIn. For the roles sourcers actually struggle with, source coverage dominates reasoning quality by orders of magnitude.

Here is the US software engineering pool, broken down the way an AI sourcing agent has to see it:

0.19% is the number to internalize. A "do what I meant" agent that only sees LinkedIn still has to find needles whose primary professional footprint is a crates.io handle, three merged PRs on tokio, and a 2024 RustConf talk. None of that lives on a LinkedIn headline.

0.19%

Why reasoning does not rescue a smaller corpus

Better intent parsing on a smaller corpus still returns a smaller corpus. If HA2 reads your prompt perfectly and the right 400 engineers are not on LinkedIn, or have a two-year-stale headline that says "Software Engineer" with no mention of Rust, the agent returns nothing, confidently and conversationally. That is worse than returning nothing bluntly, because it closes the loop in your head: you asked, it answered, you move on.

Noon, one of the newer cross-source vendors, puts the structural ceiling as sharply as anyone has: "LinkedIn reaches less than half the engineering talent market." Whether the exact number is 50% or 60%, the direction is the one that matters. HA2 is a better agent on the half you already had.

The free November upgrade is a distribution event

The important thing about the November auto-upgrade is not that HA2 is good. It is that overnight, the sourcing floor for a majority of English-speaking TA teams becomes "whatever LinkedIn can see." Workflows get rebuilt around HA2's screening and interview-coordination hooks because they are free and already wired in. The gap between what the agent reaches and what the open web contains becomes the gap between what most recruiting orgs see and what the market actually looks like.

Three second-order effects worth planning for:

1. Memory is a lock-in play. The agent gets more useful the longer you feed it. Switching costs now compound inside LinkedIn the same way they compound inside an ATS. The sourcing-quality pitch is also a vendor-stickiness pitch, and procurement should price that in.
2. Hiring managers will trust it more than it deserves. A conversational agent that writes clean summaries reads like a research analyst. For roles where the LinkedIn corpus is thin, it is a research analyst with half the sources missing and no footnotes.
3. The 64% trust stat cuts against LinkedIn. If 64% of US TA pros say it is harder to know who to trust, the fix is cross-referencing LinkedIn claims against GitHub commits, Google Scholar citations, patents, and conference talks. HA2 cannot do that cross-reference. It is a trust-signal consumer, not a trust-signal producer.

Where HA2 fits in a 2026 sourcing stack

Treat Hiring Assistant 2 as the LinkedIn-side half of a two-engine stack, not a replacement for cross-source tools. It is excellent at what it does; what it does is bounded by one database.

Here is the honest division of labor:

• Use HA2 for: high-volume, LinkedIn-native roles (enterprise AEs, CSMs, PMs at BigCo, finance leads, generalist marketing), InMail orchestration, interview coordination, pipeline hygiene, and the screening workflows LinkedIn has now wrapped the agent around.
• Do not use HA2 for: systems engineers, infra, security researchers, ML research scientists, open-source maintainers, protocol engineers, game engine devs, hardware, or anyone whose public identity is a GitHub profile and a conference talk list.
• Pair it with: a cross-source engine that reaches GitHub, Stack Overflow, Google Scholar, crates.io, and the long tail of personal sites. HireEZ, SeekOut, and AmazingHiring exist precisely because the second bucket is where competitive hires are won.

The buyer's guides already name the pattern. HireEZ aggregates candidate data from 45+ platforms including LinkedIn, GitHub, and Stack Overflow, and its AI Boolean builder auto-generates search strings for recruiters who do not know Boolean syntax. SeekOut and AmazingHiring are called out specifically because both aggregate developer data from GitHub, Stack Overflow, and code repositories. The entire category exists because LinkedIn-only sourcing has a ceiling, and HA2 does not raise it.

Price, roughly, so you can argue the stack internally

LinkedIn Recruiter Lite starts around $1,600 per year; Recruiter Corporate starts around $10,800 per year, and HA2 sits inside the Hiring Assistant SKU on top of that. HireEZ is quote-based with public listings around $169 to $250 per seat per month. The point is not which is cheaper. The point is that a cross-source seat is a line item roughly the size of a Recruiter seat, and it covers the half of the market the Recruiter seat cannot see. If you are running an engineering-heavy pipeline on one engine, you are overpaying per hire, not saving.

The Rust example, in detail

The scatter pattern inside Refolk's 1,087 US Rust engineers is the clearest argument against LinkedIn-only sourcing for technical roles.

The top employer in a 25-profile sample is Google with three engineers. After that, the distribution falls off a cliff: Oxide Computer Company with two, then Figure, Shopify, Uniswap, Helius, Udemy, Reclaim.ai, Freshpaint, and Kapwing each appearing once. That is a long-tail distribution, not a concentration.

In-network "people like X at companies like Y" lookalike search degrades fast on a distribution like that. Oxide, specifically, is a company whose hiring brand is built on public engineering writing rather than LinkedIn posts. An agent that only reads LinkedIn sees a hardware startup. An agent that reads the open web sees a magnet for a specific kind of systems-minded Rust engineer, and knows which three competitors they will respond to.

This is the exact gap Refolk closes. You describe the person in plain English - "Rust engineers who have contributed to tokio or hyper and worked at a small, rigorous infra company" - and get back a ranked shortlist pulled from GitHub, LinkedIn, and the open web as one query, not three.

A better agent on the wrong half of the market is still the wrong half of the market.

What to actually do before November

Decide, before the free upgrade lands, which of your reqs are LinkedIn-native and which are not. Do not let a free feature decide your pipeline architecture for you.

A practical pre-November checklist:

1. Audit your open reqs. Mark each as LinkedIn-native, mixed, or open-web-native. For most engineering orgs, the mixed and open-web-native buckets together are more than half the pipeline by headcount, and most of the pipeline by cost-per-hire.
2. Decide what HA2 owns. The screening, interview coordination, and InMail orchestration workflows are a real productivity win. Let it own them. Do not let it own first-pass sourcing on any role where the primary signal lives outside LinkedIn.
3. Benchmark one cross-source tool on your three hardest reqs. Pick the ones your LinkedIn sourcer has quietly stopped sending you candidates for. If a cross-source engine returns 40 real names on a req that has been dry for six weeks, the stack math writes itself.
4. Price the switching cost of HA2 memory now, not in 2027. Every requisition you feed it compounds the lock-in. That is fine if you are deliberate about it. It is a problem if you drift into it.
5. Build a cross-reference step into your trust workflow. For every HA2-sourced candidate above a certain level, require a GitHub, Scholar, patent, or conference-talk check before the hiring manager loop. That is the 64% trust stat, operationalized.

Hiring Assistant 2 is the best in-network sourcing agent LinkedIn has ever shipped. That is the right thing to say about it, and it is also the whole limitation. The half of the engineering market that lives on GitHub, in Discords, on personal sites, and in conference lineups did not get easier to find on September 29. It got easier to miss, because the agent that misses them is about to get a free upgrade and a lot more trust.

FAQ

Is LinkedIn Hiring Assistant 2 worth paying for?

If you are already a LinkedIn Recruiter Corporate customer running LinkedIn-native roles (enterprise sales, CS, PM, generalist marketing, finance), HA2 is a strong productivity layer and the November upgrade is free for English-speaking Hiring Assistant customers. The honest answer is that it is not a sourcing-coverage upgrade. It is a workflow and reasoning upgrade on the same corpus. Price it as the second half, not the first.

What are the best LinkedIn Recruiter alternatives for technical roles in 2026?

For GitHub-dense roles, the serious options are cross-source engines: HireEZ (45+ platforms including GitHub and Stack Overflow), SeekOut, and AmazingHiring (both specifically called out for GitHub, Stack Overflow, and code repo aggregation). The pattern is the same across all of them: candidates who do not keep their LinkedIn current still leave footprints across the rest of the open web, and the tool pulls those footprints into one ranked view.

Does Hiring Assistant 2 search GitHub?

No. Hiring Assistant 2 operates inside LinkedIn's data and workflow surface. It does not index GitHub, Stack Overflow, Google Scholar, crates.io, personal sites, or conference talk lineups. If your target candidate's strongest professional signal lives on any of those, HA2 will not find it, no matter how well it parses your prompt.

How do AI sourcing agents compare on source coverage?

The spread is wide. HA2 covers the 1B+ LinkedIn member base and nothing else. Cross-source aggregators like HireEZ pull from 45+ platforms including LinkedIn, GitHub, Stack Overflow, and Google Scholar. Technical-specialist tools like SeekOut and AmazingHiring lean heavily on developer-specific sources. For any 2026 AI recruiter tools decision, source coverage is the first filter and reasoning quality is the second, not the other way around.