Technology & People — May 9, 2026
Airbnb CEO: AI Lets One Engineer Do the Work of 20
Brian Chesky revealed that AI agents now write 60% of Airbnb’s new code, and that this productivity shift effectively lets one engineer produce what used to require 20 people. The company has been flattening its engineering teams accordingly. One Airbnb engineer wrote publicly that they produce “99% of their code with LLMs” and called production-quality AI code generation “a solved problem.”
This isn’t just about efficiency — it’s about team structure. When one engineer can handle what a squad used to do, the entire hierarchy of software development changes. Middle management, code review processes, deployment pipelines — all of it gets redesigned around a new assumption about individual throughput.
The human angle: For experienced engineers, this is liberation from boilerplate. For junior developers trying to break in, it’s a closed door. The companies that figure out how to build the junior pipeline in an AI-native world will have a structural advantage. The ones that don’t will find themselves with a generation of engineers who never learned to debug anything an AI couldn’t fix. — Benzinga
Salesforce Commits to 1,000 AI-Native Grads — But What Does “AI-Native” Mean?
Salesforce’s Builder program is recruiting 1,000 graduates who grew up with AI and already operate in it natively. The numbers are striking: AI-native graduates are 4x more likely to use AI daily, deliver 3x faster, and drive a 40% increase in quality of work. Chief People Officer Nathalie Scardino says these graduates “aren’t threatened by AI — they’re the ones building it.”
The “3As framework” — Attract, Assess, Activate — is Salesforce’s playbook for other businesses to copy. The implicit argument: traditional hiring pipelines are broken for an AI world, and companies need to rethink how they identify, evaluate, and onboard talent.
The human angle: “AI-native” is a generational divide in real time. Someone graduating in 2026 has never worked in a world without ChatGPT. Their expectations about what work looks like, what tools are available, and what “productivity” means are fundamentally different from someone who entered the workforce in 2019. That gap is going to create friction — and opportunity. — Salesforce News
NZ Businesses Losing $2.2M Average to AI Identity Fraud
AI-generated synthetic identities and deepfakes are hitting New Zealand businesses hard. The average cost of AI-generated identity fraud per business has reached $2.2 million, driven by sophisticated attacks that bypass traditional KYC verification. Globally, synthetic identity fraud is the fastest-growing financial crime, with AI generating fake passports, driver’s licenses, and even live video deepfakes for verification calls.
Recent incidents include a Dutch bank where deepfakes opened 46 fraudulent accounts, and an Indian case where AI bypassed Aadhaar biometric security to secure fraudulent loans. The OECD has catalogued these as active, escalating threats.
The human angle: This is the dark side of the AI productivity story. The same generative AI that writes code and creates art also generates perfect fake IDs. And the victims aren’t just banks — they’re individuals whose identities get stolen, small businesses that get defrauded, and eventually, all of us who will face more friction verifying who we are online. — Newsroom NZ | OECD AI Incidents | IProov
NZ’s AI Data Centre Boom: Who Actually Benefits?
The Conversation NZ published a pointed analysis asking exactly who benefits from New Zealand’s AI data centre build-out. Datagrid’s 280MW Southland facility will be the second-largest electricity user in the country — equivalent to a major aluminium smelter — but will create relatively few permanent jobs. The power consumption alone raises questions about electricity pricing for households and small businesses.
The NZ Parliamentary committee report adds that most international businesses aren’t seeing AI returns, and 90% of surveyed firms saw no productivity impact. So New Zealand is building AI infrastructure — consuming a huge share of its clean energy — for an industry that hasn’t proven its economic value yet.
The human angle: There’s a colonial echo here. New Zealand exporting raw materials (in this case, clean energy and stable geography) while someone else captures the value. Data centres are the new dairy farms — big installations that consume local resources while profits flow overseas. The question is whether we can build enough local AI capability to change that equation. — The Conversation | Newsroom NZ
The Mythos Aftermath: How One Incident Reshaped AI Oversight
The Mythos crisis — the hacking AI incident that shocked Washington — has permanently reshaped AI governance. Google, Microsoft, and xAI have now joined OpenAI and Anthropic in submitting to pre-release government testing at the Commerce Department’s CAISI office. The voluntary agreements were catalysed directly by the national security panic that followed Mythos.
The centre has conducted 40+ evaluations, testing for biological weapon synthesis, cyberattack automation, and autonomous agent behaviours. The government has zero statutory power to block a release. It’s all voluntary. And it all exists because of one incident that demonstrated what a sufficiently capable AI model could do.
The human angle: One event reshaped the entire AI oversight landscape. The lesson is uncomfortable: it took a genuine near-miss to get five competing companies to voluntarily submit to government evaluation. The Mythos crisis didn’t just change policy — it changed the industry’s relationship with risk. Companies that previously competed on capability now also compete on safety credibility. — The Next Web | Bloomberg
The AI Productivity Paradox: 90% of Firms See No Impact
The NZ Parliamentary committee report cites a National Bureau of Economic Research study finding that 90% of surveyed firms saw no impact of AI on workplace productivity. Victoria University AI expert Andrew Lensen says aggressive AI uptake hasn’t led to quick wins, and that much of the adoption wave consists of companies “simply enabling Copilot, declaring they are AI-enabled, and calling it a day.”
The comparison to the dot-com bubble is raised by some economic modelling — massive investment, questionable returns, but eventual transformation. The question is whether we’re in 1999 or 2003.
The human angle: This is the most important story in AI that nobody wants to talk about. The hype cycle is running ahead of the reality. For the average knowledge worker, AI tools are useful but not transformative. For the average business, the ROI calculations don’t add up yet. That doesn’t mean it won’t happen — but it means we’re living through the “trough of disillusionment” phase, and it’s going to feel real for a while. — Newsroom NZ
🔍 THE BOTTOM LINE
The technology story this week is a study in contradiction. Airbnb says one engineer can do the work of 20 with AI, Salesforce is hiring graduates who never knew a world without LLMs — and yet 90% of businesses see zero productivity gain. Both things are true. The leading edge of AI adoption is producing genuinely astonishing results. The trailing edge is confused, cautious, and unconvinced. The gap between them is where the real story lives.
❓ FAQ
Are AI-native graduates really different? Salesforce’s data says yes — 4x more likely to use AI daily, 3x faster delivery, 40% higher quality. But “AI-native” is more about mindset and comfort than specific skills.
Is AI identity fraud getting worse? Significantly. AI-generated synthetic identities, deepfakes that bypass KYC, and fake documents are the fastest-growing financial crime category. The tech is widely available and cheap.
Will AI eventually deliver productivity gains? The historical pattern from previous tech revolutions suggests yes. But the timing is uncertain — we could be years away from the kinds of productivity gains that show up in national statistics.
Got a take on the AI productivity question? We’d love to hear it.