Every month, Patryk Grzelak, Ben's Engineering Lead, rounds up the most interesting shifts he’s seen in AI. This helps us spot the trends that hint at what’s coming next for the way we work.
This month’s biggest AI stories weren’t about headline-grabbing demos - they were about infrastructure, access, and what happens when tools get just good enough to be quietly transformational.
Meta offers $100M signing bonuses to AI researchers
TL;DR: Meta is reportedly offering OpenAI researchers up to $100 million to jump ship.
AI talent is rare. Fewer than 1,000 people globally are considered top-tier AI researchers. Meta, once a frontrunner in AI, has been playing catch-up - and now it’s pulling out all the stops to rebuild its research team.
Meanwhile, DeepMind is quietly updating employment contracts to reinforce non-competes and reduce attrition. The AI talent war isn’t just alive - it’s escalating.
Why this matters for HR:
It’s a wake-up call about retention and workforce planning - not just for AI teams, but for any business competing globally for specialised skills. In HR, this raises important questions:
- Are your high-value roles insulated from headhunting risk?
- Can you flex compensation, equity, or benefits packages quickly if needed?
- Do your long-term talent strategies account for rising cross-industry competition?
For global companies, managing retention at scale - and across markets - is becoming a strategic imperative.
Claude access cut after OpenAI acquisition
TL;DR: Anthropic cut API access to a startup OpenAI had just acquired.
After OpenAI bought Windsurf - a startup building with multiple AI models - Anthropic (maker of Claude) immediately cut off their API access. The reason? Anthropic didn’t want OpenAI to indirectly benefit from its models.
Why this matters for HR:
It’s a signal that AI collaboration is becoming more protective and closed. Providers are starting to treat models like proprietary tech, not open research tools.
If you’re evaluating vendors or building internal tools that depend on third-party AI, this raises two flags:
- Stability of access - Can your tools operate if an AI provider changes terms or blocks access?
- Compliance and privacy - Do you know where your employee data flows, and who has control?
For HR leaders responsible for tech-enabled benefit platforms, automation, or internal search tools - make sure any vendors or technology partners you work with have backup systems or alternative solutions ready if their primary AI provider cuts off access. Especially if you're working across multiple regions with different compliance standards.
OpenAI cuts prices by 80% and launches O3 Pro
TL;DR: The cost of using top-tier AI just dropped significantly, making it far more accessible.
OpenAI’s previous flagship model (O3) is now 80% cheaper to run. Alongside this, they’ve released a more advanced version - O3 Pro - which offers improved accuracy and real-time web access, albeit with slower response times.
The real story isn’t just cost - it’s capability growth. AI is becoming more affordable and more powerful at the same time.
Why this matters for HR:
For enterprise HR teams managing global workforces, this opens the door to scalable AI tools that were previously cost-prohibitive - think personalised onboarding flows, smarter knowledge search, or multilingual policy assistance.
Cheaper usage-based pricing means these tools can now be tested and rolled out across regions without blowing through budgets. It's a significant shift in how accessible AI-driven HR tech becomes for large and distributed teams.
O3 Pro and the rise of the ‘Gentle Singularity’
TL;DR: OpenAI’s CEO predicts near-term breakthroughs - not in hype, but in quiet competence.
Alongside the launch of O3 Pro, OpenAI CEO Sam Altman published a thoughtful essay outlining a vision for the next two years:
- In 2026, we may see AI systems generate entirely new insights.
- In 2027, we may see useful physical robots.
- Experts who embrace AI will still outperform those who don’t.
O3 Pro isn’t flashy - it’s slow, precise, and grounded in real-world sources. But that’s exactly what makes it interesting: this is AI designed for serious, background work.
Why this matters for HR:
Strategic HR teams are already using AI to streamline processes and decision-making. This next generation of AI could enable:
- More context-aware, localised benefits communication
- Insight-driven benchmarking based on real-time industry data
- Automation of high-effort, low-impact admin - like eligibility tracking or document processing
The takeaway? You don’t need ‘shiny’ AI to drive impact. The most useful tools will likely be the ones that run quietly - but reliably - behind the scenes.
Agents that work while you sleep
TL;DR: Startups like Cursor are giving developers background AI agents that ship code overnight.
Cursor, a developer platform, now lets engineers assign AI agents to make updates to live codebases while they’re offline. It builds on their existing work and follows instructions, essentially giving teams a productive night shift.
While it’s early days, this shift toward autonomous task agents is gaining momentum.
Why this matters for HR:
As these agents become more widely applicable, HR teams should start considering how they could be used to:
- Draft policies or documentation overnight
- Generate employee communications in local languages
- Analyse anonymised feedback at scale
For globally distributed teams, especially those working across time zones, these agents offer a way to keep work progressing asynchronously - even while your teams sleep.
The real risk to jobs might be the headlines
TL;DR: Industry leaders are pushing back on the narrative that AI will cause mass job loss.
Nvidia CEO Jensen Huang recently criticised public claims that AI will eliminate swathes of jobs. He’s not alone - OpenAI and other major players are taking a similar stance: AI will change jobs, not replace them.
There’s even a leaderboard now tracking which startups are delivering the most output with the leanest teams: Lean AI Leaderboard.
Meanwhile, some companies are still leaning hard on 'AI will replace you' messaging. One startup, Artisan, is running adverts promoting AI salespeople. The ‘Era of AI Employees’ might arrive some day, but not today. Just this weekend, I was using the latest Gemini models which were just released; which accidentally leaked some internal Google data!
%202.png)
Artisan relies on fear/hope marketing to push a product that helps sales teams find and engage new prospects. If you know anything about outbound sales email marketing, you'll know you can tell a ChatGPT-created outbound email from a mile away. It may be bold marketing, but it's not particularly helpful or realistic.
Why this matters for HR:
HR leaders have a responsibility to set the tone - both internally and externally. That means:
- Framing AI as augmentation, not elimination
- Investing in re-skilling and capability-building now
- Supporting managers through the shift, especially in frontline or operational roles
For enterprise orgs, the narrative you build around AI adoption will shape employee engagement and retention just as much as the tech itself.
{{cta="/snippet/patryk-ai-blog-july-cta"}}