AI is creating more work, at least for now
For all the talk of AI wiping out jobs, the data is telling a calmer story. A new Vanguard analysis suggests that roles most exposed to AI are actually seeing stronger wage and job growth than those with little exposure.
Vanguard compared occupations where AI can actively support work, like software development and data analysis, with roles where it has limited impact, such as construction or cleaning. Over the past two years, wages in high AI exposure jobs rose nearly four times faster, and employment growth also pulled ahead.
That does not mean the labor market is thriving. Hiring has slowed, especially in tech and professional services. But economists say those pressures have more to do with higher interest rates, federal job cuts, and tighter immigration rules than automation replacing workers.
In some cases, AI is making work cheaper and faster, which increases demand. Box CEO Aaron Levie argues that lowering the cost of building software expands the number of problems companies can afford to solve, which ultimately creates more work rather than less.
None of this rules out future disruption. The AI transition is still early, and infrastructure spending is running far ahead of hiring. But for now, the evidence cuts against the idea that AI is already hollowing out the job market.
Supported by HubSpot
Tired of disconnected tools? Same. Make 2026 the year you reduce the chaos and automate what slows you down. Discover top apps that pair well with HubSpot. So you can do less busywork and more selling. Focus on the work that matters.
OpenAI explores a $10B+ Amazon deal as compute costs climb
OpenAI is in talks with Amazon about a potential investment of at least $10 billion, a deal that could value the company above $500 billion and deepen its dependence on AWS.
The funding would help cover massive compute commitments, including OpenAI’s recently announced $38 billion spend with AWS over the next seven years. In return, OpenAI would begin using Amazon’s Trainium chips, giving Amazon a high profile customer for its Nvidia rival hardware.
The discussions also touch on a possible commerce partnership. OpenAI wants ChatGPT to evolve into a place where users complete tasks and make purchases, while Amazon is looking for ways to stay close to the AI stack even as Microsoft retains exclusive rights to sell OpenAI models through Azure.
The talks are still fluid, but the direction is clear. OpenAI’s growth now depends on unprecedented levels of capital and compute, and Amazon wants a bigger role in how that infrastructure gets built.
Cisco lays out a clearer playbook for AI security
Cisco is rolling out a new framework aimed at helping companies identify and manage risks that are specific to AI systems.
Its Integrated AI Security and Safety Framework maps nearly 20 categories of threats, including prompt injection, jailbreaking, multimodal attacks, and training data poisoning. Cisco argues that most existing security models were built for traditional software and miss how AI systems fail in practice.
The goal is to give security teams and executives a shared language for spotting risks and prioritizing defenses. Cisco is aligning its own AI Defense tools with the framework and says it plans to work with standards bodies to encourage broader adoption.
As AI moves deeper into core business workflows, Cisco is betting that clearer threat models will be essential to keeping systems reliable and safe.
GPT-5 shows early signs of doing real lab work
GPT-5 has demonstrated it can contribute to hands on laboratory research, a step toward AI playing a more direct role in scientific discovery.
In a controlled collaboration with biosecurity startup Red Queen Bio, OpenAI tested whether the model could improve wet lab protocols rather than just analyze data. GPT-5 proposed changes to a standard molecular cloning process, researchers ran the experiments, and the model refined its approach based on the results.
The outcome was a 79x efficiency improvement. The researchers say the model went beyond simply remixing published work and showed early signs of creative problem solving.
The results are early and tightly scoped, and OpenAI is careful not to frame them as a breakthrough. But the experiment points to a future where AI helps scientists iterate faster, reduce costs, and explore ideas that would take humans much longer to test.
Google brings its vibe coding tool Opal into Gemini
Google is integrating its vibe coding tool, Opal, directly into the Gemini web app, making it easier to build small AI powered apps without writing code.
Opal now lives inside Gemini’s Gems manager, where users can describe what they want in plain language and watch Gemini turn it into a step by step workflow. A visual editor lets users rearrange actions, connect steps, and see how the app works under the hood.
Google has also added a view that converts written prompts into structured logic, making the process more transparent. More advanced builds can be pushed into Opal’s standalone editor.
The move brings Google closer to rivals like OpenAI and Anthropic, as well as startups such as Cursor and Lovable, as vibe coding tools continue to spread beyond developers to everyday users.
<img src=’https://www.vpdae.com/open/c8b662c0.gif?opens=1’ width=’1’ height=’1’>