The last 7 days were a reminder that AI hiring is now constrained by three things: compute capacity, security risk, and agent reliability. OpenAI says it has already surpassed its original 10GW US infrastructure-by-2029 Stargate milestone, with 3GW added in the last 90 days. At the same time, Anthropic’s Mythos program (built to find thousands of zero-days) is now dealing with reports of unauthorized access via a third party, which is exactly the kind of event that makes boards suddenly fund security headcount. And on the product side, Adobe has started public testing an agentic Firefly assistant that can run workflows across multiple Adobe apps, and is even building a lighter version to plug into external chatbots starting with Claude.

The Drop

1) OpenAI says Stargate already cleared its 10GW milestone

What happened: OpenAI says that when it announced Stargate (Jan 2025) it committed to securing 10GW of US AI infrastructure by 2029, and that just over a year later it has already surpassed that milestone, with 3GW added in the last 90 days.
Why it matters for hiring: this is the clearest possible signal that the limiting factor is not ideas or models, it’s power, buildout, and operations.
Roles likely to spike:

• Data center infrastructure, networking, storage, SRE

• Capacity planning + FinOps (cost-per-token, utilization, forecasting)

• Energy and site strategy (power procurement, grid constraints)

2) OpenAI loosens Microsoft ties as it widens partners

What happened: Axios reports Microsoft relinquished its exclusive rights to sell access to OpenAI’s models, while still remaining the primary cloud provider. OpenAI is also exploring a hardware partnership with Qualcomm, per the same report.
Why it matters for hiring: distribution and infrastructure strategy are now product decisions. The teams winning will be the ones that can ship across clouds, manage model portability, and commercialize partnerships without breaking delivery.
Roles likely to spike:

• Multi-cloud platform engineers

• Partner engineering and solutions engineering

• Hardware-adjacent product and infra roles (if the Qualcomm track moves)

3) Mythos and Glasswing: AI security just got a new baseline

What happened: Anthropic’s Glasswing page says Claude Mythos Preview identified thousands of zero-day vulnerabilities across every major OS and browser. TechRadar reports Mythos was accessed by unauthorized users via a third-party evaluator environment (citing Bloomberg), and notes Mozilla’s CTO had previously reported Mythos finding 271 vulnerabilities in a Firefox build.
Why it matters for hiring: this is the moment a lot of companies stop treating AI security as “later” and start treating it as “budget now.”
Roles likely to spike:

• AI security engineers and red team automation

• Vulnerability research + AppSec

• SOC automation and detection engineering

• Security engineering for vendor and evaluator pipelines

4) Adobe starts public testing agentic Firefly (and is wiring it into Claude)

What happened: Adobe began public testing an agentic AI assistant inside Firefly that can execute multi-step workflows across Adobe apps. It is also developing a lighter version designed to integrate into external chatbots, starting with Anthropic’s Claude.
Why it matters for hiring: “agentic UX” is becoming the product surface. That pulls hiring toward orchestration, reliability, permissions, and evaluation, not just model calls.
Roles likely to spike:

• Agent workflow engineers (tool-use, retries, state)

• Product engineers who can ship cross-app automation safely

• Trust, safety, and compliance roles for action-taking assistants

5) Defense and space hiring stays hot: Pentagon AI deals + True Anomaly raises $650M

What happened: The Guardian reports the Pentagon inked deals with seven AI companies (SpaceX, OpenAI, Google, Nvidia, Reflection, Microsoft, AWS) for classified military work.
Separately, True Anomaly announced a $650M Series D, taking it to $1B+ total capital raised. (TNW reports the round at a $2.2B valuation.)
Why it matters for hiring: defense AI and space security keep pulling from the same pools as infra and robotics.
Roles likely to spike:

• Systems, simulation, autonomy, and security engineering

• Cleared roles and gov-facing delivery teams

• Platform reliability for real-world deployments

AI Tool of the Week (new pick)

HackerRank AI Assistant (Guarded vs Unguarded interviews)

What it does: HackerRank’s AI Assistant supports interviews in modes like Unguarded (real-world AI tool usage) and Guarded (more constrained help), letting you test how candidates work with AI rather than pretending it doesn’t exist.
Who it’s for: teams hiring engineers in 2026 who want signal on judgment, validation, and AI leverage during real work.
Quick pilot idea (this week):

• Pick one role and run 10 interviews with AI Assistant enabled

• Use Unguarded for senior roles (judgment signal), Guarded for junior (baseline competence)

• Add one scoring line item: “AI leverage + validation discipline”
  Metrics to track:

• Pass-through rate vs your current format

• Time-to-decision (does feedback get clearer?)

• False-positive reduction (fewer “great talker” hires)

Hiring / Interview Insight

“AI leverage” is now measurable, so measure it

We’ve got a market full of supply, but hiring is still hard because companies are selecting for different traits than two years ago. TrueUp’s tracker shows 262 tech layoffs impacting 119,625 people so far in 2026.
If you want to win in this environment, stop guessing: add one station that explicitly tests how someone uses AI under constraints (security, correctness, speed). Use the HackerRank pilot above as the easiest way to do this without reinventing your loop.

Funding Watch

• True Anomaly | $650M Series D | $1B+ total raised 
  Likely hires: systems, security, autonomy, simulation, program delivery.

• Netomi | $110M Series C | $160M+ total raised | ~170 employees 
  Likely hires: forward-deployed / enterprise implementation, agentic CX engineering, integrations.

• Legora | $50M extension | $600M Series D total | $5.6B valuation 
  Likely hires: applied AI, legal workflow product, enterprise rollout, data quality and evals.

(If you want, next week we can add a “roles likely to hire first” line for each round in the intro so it’s even more scannable.)

Quick Bytes

• Reuters reports Anthropic is weighing a new round that could value it at $900B+ (early-stage discussions, not accepted offers).

• Hundreds of contractor workers training Meta’s AI systems may be cut at a vendor in Ireland, per Wired.

• Wizards of the Coast (Magic: The Gathering Online) devs are unionizing, citing layoffs, generative AI usage, and remote work issues.

• Itron confirmed a cyberattack with internal systems accessed (detected April 13) per its SEC filing and TechRadar coverage.

What to do this week

1) Add a capacity owner to your AI roadmap

Metric: utilization, cost per inference unit, lead time for new capacity.

2) Run a third-party evaluator and vendor access audit

Mythos access via third parties is the warning shot.
Metric: privileged accounts reduced, MFA coverage, time-to-revoke, vendor access inventory completeness.

3) Pilot one “AI leverage” interview change

Use HackerRank AI Assistant for 10 interviews and compare signal quality.
Metric: pass-through stability + 30-day hiring manager satisfaction.

This week’s pattern is simple: compute is scaling, agents are becoming the workflow, and security is the tax you pay to participate. If your hiring plan is still “just hire more engineers,” you’re behind. The winners are hiring infra, security, and people who can operate AI systems safely at scale.

That’s all for this week’s Tech Talent Drop — stay informed, and see you next week!