AI Hiring’s Cheap Talk Problem - Xist4

August 13, 2026

AI Hiring’s Cheap Talk Problem

Last week, a tiny AI model landed a punch that should make every founder, CTO and hiring leader sit up a little straighter.

Pathway’s 150M model scored 29.5% on ARC-AGI-1 while costing 11 times less than ChatGPT’s comparable reasoning model at inference, according to TechRadar’s coverage of the story. Source: TechRadar, ‘11X cheaper than ChatGPT: Tiny 150M model just proved AI doesn't need to think out loud to be smart’.

Now, this is not an AI blog. At least not officially. But if you work in hiring, especially in tech, data, infrastructure or cyber, this should ring a very loud bell.

Because recruitment has its own version of ‘thinking out loud’. And frankly, a lot of it is expensive fluff wearing a nice blazer.

We have built far too many hiring processes that confuse performance with performance art.

More stages. More chatter. More interview rituals. More polished waffle. And somehow we act surprised when the best people get bored, disappear, or worse, join a competitor who made a decision before lunch.

Introduction

The Pathway story matters because it challenges a lazy assumption: that visible complexity equals better outcomes.

In AI, that means assuming the model that produces more reasoning tokens must be doing better thinking.

In hiring, it means assuming the candidate who gives the longest, smoothest, most consultant-friendly answer must be the strongest hire.

Not true.

Some of the best engineers, BI leaders, infrastructure specialists and cyber talent I have placed were not the loudest people in the room. They were just sharp. Clear. Effective. Low ego, high impact.

Funny how often businesses miss them while admiring someone who can talk for twelve uninterrupted minutes about ‘cross-functional stakeholder alignment’.

Very impressive. Still cannot fix your broken platform.

Hiring loves theatre more than signal

Let’s say the quiet bit out loud: most hiring processes are bloated because bloated feels safer.

If five people interview a candidate across seven stages, surely the decision must be good, right?

Not necessarily. It might just be slower, murkier and more political.

This is the corporate version of ordering the biggest tasting menu because you assume more courses means better food. Sometimes it does. Sometimes you just paid £180 to leave hungry and confused.

The same goes for hiring.

Businesses often stuff in extra interview rounds, presentation tasks and stakeholder chats because they lack confidence in what they are actually assessing.

So they compensate with process.

That usually creates three problems:

  • You test stamina, not capability. Great candidates are not trying to win a reality show.

  • You reward polish over substance. Some people interview brilliantly and execute terribly.

  • You lose the market. Top talent rarely hangs around while your team schedules Interview Round Six: The Cultural Vibe Sync.

And yes, I gave that stage a silly name. But not by much.

Expensive does not mean effective

The TechRadar piece highlights something bigger than cost reduction. It points to efficiency as a competitive edge.

That idea should be tattooed onto the inside of every hiring team’s eyelids.

Because too many firms still behave as if the pricier, more elaborate route must be the better route.

In recruitment, that usually shows up like this:

  • Using a top-heavy process for every role, regardless of seniority or scarcity

  • Asking candidates to complete unpaid tasks that tell you little about real-world impact

  • Briefing agencies badly, then blaming the market when the search drags on

  • Paying the hidden cost of vacancies while obsessing over visible hiring fees

That last one is the killer.

I have seen companies agonise over recruiter fees while happily absorbing six months of delayed delivery, team burnout, missed revenue and a manager spending half their week firefighting because the role is still empty.

That is like stepping over pounds to save pennies.

Or, for the fintech crowd, turning down a quality hire to protect budget while your roadmap quietly catches fire.

The best hiring processes are boringly clear

Here is my slightly provocative view: the best recruitment process is usually the one that feels almost underwhelming in its simplicity.

Not careless. Not rushed. Just clean.

It knows what good looks like.

It assesses against real outcomes.

It involves the right people, not every person with a calendar.

And it gets to a decision before doubt breeds nonsense.

If Pathway’s model can force a rethink on AI efficiency, leaders should do the same with hiring.

Ask yourself:

  • What are we truly trying to measure?

  • Which stage gives us the strongest signal?

  • Which stage exists mostly to make us feel thorough?

  • Are we selecting for actual performance, or confidence theatre?

If those questions make anyone uncomfortable, good. That is usually where the useful work starts.

A sharper framework for smarter hiring

When I work with clients, I push for a simple lens. I call it Signal, Speed, Sell.

Signal

Every stage should reveal something meaningful about a candidate’s ability to do the job.

Not whether they can survive poorly designed hoops.

Good signal might include:

  • A properly scoped technical discussion

  • A case study based on realistic business problems

  • Evidence of outcomes from past roles

  • Questions that test judgement, not rehearsed jargon

Weak signal includes:

  • Generic competency chats repeated by three different people

  • Massive take-home projects with no clear scoring criteria

  • Interviewers freelancing random pet questions

Speed

Speed is not recklessness. It is respect.

Top candidates interpret delay as disinterest, dysfunction, or both.

If your process takes four weeks to organise a second-stage interview, the market has moved on.

And no, the market does not care that Karen was on annual leave.

Sell

Hiring is not an interrogation. It is a two-way decision.

The strongest candidates are assessing your clarity, leadership and urgency from the first conversation.

If your process feels vague, overly political or painfully drawn out, that is not a neutral experience. It is a warning label.

The companies winning talent right now are not just better at assessment. They are better at making serious people want to join.

What this means for tech leaders right now

If you are hiring in infrastructure, cloud, BI, data, engineering or cyber, the market has changed.

The strongest candidates are more selective. Teams are leaner. Budgets are scrutinised. And tolerance for waste is lower.

That means your hiring process has to work harder with less drama.

Here is what I would do this quarter if I were auditing your process:

  • Cut one stage. If you cannot justify it clearly, remove it.

  • Define success before interviewing. What must this person achieve in 12 months?

  • Score candidates against outcomes. Not vibes. Not shared hobbies. Not who interviewed best on a Friday afternoon.

  • Brief properly. A vague brief creates vague candidates and long searches.

  • Move like you mean it. Great people are rarely available for long.

And if your team says, ‘We need more data before deciding,’ sometimes what they really mean is, ‘We are nervous about accountability.’

That is a leadership problem, not a recruitment process.

The real lesson from the tiny model

The big idea here is not that small always beats big.

It is that efficiency exposes vanity.

Pathway’s model, as reported by TechRadar, suggests that intelligence does not always need a lot of visible verbal scaffolding to prove itself. Source: TechRadar.

Hiring should learn from that.

You do not need endless process to make good decisions.

You need clarity.

You need signal.

You need conviction.

And, ideally, you need fewer meetings where six people say, ‘I quite liked them, but…’ before producing absolutely no useful evidence.

That sentence has probably delayed more hires than any skills shortage.

Conclusion

Here is the punchline: if your hiring process needs constant explanation to justify itself, it is probably too expensive, too slow and too performative.

The smartest systems, whether in AI or recruitment, are not the ones making the most noise. They are the ones delivering the best result with the least waste.

That is the bar.

So before you add another interview stage, another panel, another task, ask the only question that matters.

Is this making us smarter, or just making us feel busy?

Those are not the same thing. Not even close.

If you want to hire better across tech, data, infrastructure or cyber, stop chasing theatre. Build a process that spots signal fast, treats candidates like grown-ups and ends with a decision.

Sharp beats loud. Every time.



Back to news