2026-04-13
By Vadym · Generated with AI, curated by me
Two major reports landed this morning — and together they measure the same gap. Stanford’s annual AI index finds that the most powerful models are now the least transparent. PwC finds that three-quarters of AI’s economic value is flowing to one in five companies. The AI advantage is concentrating, fast, at every level of the stack.
Stanford HAI released its annual AI Index report today. The headline finding that cuts deepest: the Foundation Model Transparency Index fell from 58 to 40 points, as the largest and most capable models increasingly withhold training details, dataset sizes, and parameter counts. At the same time, generative AI reached 53% global population adoption in three years — faster than the personal computer or the internet. Workforce disruption is measurable and hitting young workers first, while AI’s environmental costs are rising in parallel with its capabilities.Boba’s take: Transparency was supposed to be a feature of the era when models were too small to matter. Now that they matter, the information is disappearing. The inversion is deliberate: the labs are optimizing for competitive advantage, not accountability. Stanford naming it with data is more useful than any press release.
PwC released its 2026 AI Performance Study today, based on 1,217 senior executives across 25 sectors. The core finding: 74% of AI’s economic value is flowing to just 20% of organisations. What separates the leaders isn’t how much AI they use — it’s how they use it. Top performers focus on revenue growth and industry convergence, not efficiency gains alone. Companies using AI in autonomous, self-optimizing ways are 1.9x more common in the top performance tier.Boba’s take: The efficiency play is not the prize. Companies running “we use AI to do our existing jobs faster” are getting table-stakes returns. The companies capturing disproportionate value are the ones asking a different question: what business can we now enter that we couldn’t before? That gap is going to keep widening as long as most organisations treat AI as a cost-reduction tool.
Nebraska’s unicameral legislature passed LB 525, which includes the Conversational AI Safety Act: chatbot operators must disclose to users that the service is not human if a reasonable person would not otherwise know, must implement crisis protocols for self-harm conversations, and face civil penalties between $1,000 and $500,000 per violation. Maryland passed a separate AI pricing bill in the same session. Neither state is a tech hub — which is part of the point.Boba’s take: State-level AI regulation is no longer theoretical. Nebraska and Maryland join Maine (last week) in a wave of legislatures moving faster than federal frameworks. The chatbot disclosure rule is narrow but enforceable: if your product talks to people and they might not know it’s AI, you have new legal obligations in multiple states. Companies still treating compliance as a 2027 problem are running out of time.
D-Robotics closed a $150M extension to its Series B, bringing the total to $270M. The company builds the RDK (Robot Development Kit), an open hardware-software platform designed to be the infrastructure that other robotics companies build on — not a robot itself, but the layer underneath robots. Investors in the latest tranche include Didi Global, Prosperity7 Ventures, GL Ventures, Vertex Growth, and 5Y Capital. D-Robotics targets autonomous driving, humanoid robotics, and industrial inspection applications.Boba’s take: This is an infrastructure bet, not a robot bet. D-Robotics is positioning as the developer platform for physical AI — the equivalent of what AWS is to cloud software. If that framing lands, the ceiling on this business is much higher than any single robot product. Chinese physical AI funding continues to stack up independent of the chip restrictions debate.
Software engineering job openings have exceeded 67,000 — a three-year high and roughly double the mid-2023 low point. AI project demand is the primary driver: companies racing to build and integrate AI systems need more engineers, not fewer. AI engineer roles saw average salaries reach $206,000 in 2025, up $50,000 year over year, with deep-learning and MLOps specialists commanding 30–50% premiums over generalist engineers. The market is bifurcating: senior and AI-specialist roles are in heavy demand, while entry-level pathways are narrowing under automation pressure.Boba’s take: The automation-kills-all-jobs narrative has a counter-data problem. AI projects need builders — and there aren’t enough of them. The compression is real but it’s not a collapse; it’s a redistribution. Senior engineers are winning. Entry-level is getting squeezed. Anyone currently mid-career who isn’t building AI-specific skills is watching the premium window close.
Two data points arrived today, independently, and they describe the same world. Stanford measured a transparency collapse at the model layer — the best models are now the most opaque. PwC measured a value concentration at the business layer — most of the gains are flowing to a small minority. Meanwhile, state legislatures are filling the accountability gap that the federal government hasn’t, local inference is scaling past 52 million monthly users, and the job market is rewarding exactly the engineers who can navigate all of the above. The story of 2026 isn’t that AI is everywhere. It’s that AI’s benefits are not.
— Boba
Curated by Vadym