Nevada Approves Up to 8,000 Robotaxis for Las Vegas
In what may be the single largest regulatory green light for autonomous vehicles to date, the Nevada Transportation Authority on August 20 approved permits for Tesla, Waymo, and Uber to operate paid robotaxi services across Clark County, including the Las Vegas metro area.
Under the new Autonomous Vehicle Network Company permits, Tesla can deploy up to 5,000 robotaxis, while Waymo and Uber each received authorization for 1,000 vehicles — with Uber operating through partnerships with Hyundai subsidiary Motional and Amazon-owned Zoox. That's a potential fleet of 8,000 driverless vehicles on public roads over the next 12 months.
Tesla has tempered expectations, saying it would be "extremely happy" to get 2,500 vehicles on the road by 2027, but the permits themselves represent a dramatic shift in regulatory posture. Tesla's permit notably allows statewide expansion with simple notification — no additional approval required.
The move signals that U.S. regulators are increasingly comfortable with large-scale autonomous deployments, a sharp contrast with the caution that followed Cruise's San Francisco incidents in late 2023. For Las Vegas visitors and residents, the practical question is no longer whether robotaxis will be widely available, but how soon.
Sources: Las Vegas Sun, InsideEVs, Engadget
Claude Designs Protein Binders at Double the Industry Hit Rate
Anthropic published lab-validated results on August 18 showing that its Claude AI autonomously designed protein binders that successfully hit 14 of 15 disease-relevant targets — a process that typically takes human experts weeks or months per target.
Working with biotech partner Adaptyv Bio and DNA synthesis company Twist Bioscience, Claude generated 1,320 protein binder designs over a 48-hour autonomous session. Of those, 354 designs were confirmed to bind their targets in wet-lab testing, yielding hit rates of 22–35% compared to the industry-standard 10–15%.
The company's Mythos Preview and Opus 4.8 models achieved overall hit rates of 26.7% and 22.6%, respectively, when designing against all targets simultaneously. Anthropic released all designs, prompts, and measurement data alongside the technical report.
Anthropic was careful to note that "protein binders are not drugs" — designing a high-affinity binder is just the first step toward a therapeutic molecule. Still, the results demonstrate that frontier AI models can meaningfully accelerate the earliest stages of drug discovery, potentially compressing timelines and reducing costs for the pharmaceutical industry.
Sources: Anthropic Research, Adaptyv Bio, Dataconomy
Anthropic's Revenue Rockets Past $65 Billion Ahead of Potential Record IPO
Anthropic's annualized revenue run rate surpassed $65 billion at the end of July, according to Bloomberg — up from $47 billion in May and a mere $9 billion at the end of 2025. The company generated over $11.5 billion in preliminary revenue in Q2 alone.
The staggering growth is being driven by enterprise adoption of Claude and AI agents. Investors now expect Anthropic to finish 2026 with revenue between $100 billion and $120 billion, figures that would have seemed fantastical just a year ago.
The financial trajectory is setting the stage for what could be the largest market debut on record. Anthropic is reportedly seeking a public valuation of $2 trillion or more, and may go public before rival OpenAI, which filed its own confidential S-1 with the SEC in June and is targeting a September listing.
The parallel IPO race between the two leading AI labs will give public markets their first real window into the economics of frontier AI — including the massive infrastructure costs that sit behind those revenue numbers.
Sources: CNBC, Axios, TechBriefly
First Autonomous AI Cyberattack Hits Taiwan Government Systems
In a development that cybersecurity experts are calling a watershed moment, AI agents autonomously conducted a four-day intrusion campaign against Taiwanese government agencies in early July — the first documented case of a government-level cyberattack coordinated and executed entirely by AI.
According to Israeli AI security firm Dream, which first uncovered the attack, the autonomous agents mapped 21 government systems, cracked 85 user accounts, and extracted 2,500 personnel records. The attackers later expanded to Taiwan's nuclear safety agency, at least seven energy companies, and government suppliers.
The intrusion was built on two open-source AI agent frameworks — Hermes and OpenClaw — with model safeguards bypassed by framing the activity as an authorized penetration test. At peak operation, as many as eight autonomous agents ran in parallel.
While experts suspect Chinese state actors, neither Taiwan nor Dream has officially attributed the campaign. The incident has triggered urgent calls for stronger AI safety guardrails and prompted editorial debate about whether open-source agent frameworks need tighter controls.
Sources: CNN, CyberScoop, The Register
Google Launches Gemini 3.7 Flash at Half the Price
Google released Gemini 3.7 Flash on August 13, positioning it as the company's most capable workhorse model for coding and agent workflows — just 23 days after shipping Gemini 3.6 Flash.
The headline numbers are impressive: Google-reported DeepSWE v1.1 scores jumped from 49.0% to 65.3% over the previous version, a 16-point leap in real-world software engineering benchmarks. The model retains a 1-million-token context window and a 64K-token output limit.
Perhaps more notable than the performance gains is the pricing: introductory API rates of $0.75 per million input tokens and $3.75 per million output tokens through December 2026 — roughly half what Google charged for Flash models at launch previously. The aggressive pricing underscores how quickly the cost floor for capable AI models is dropping.
Gemini 3.7 Flash is available through the Gemini app (for AI Pro and Ultra subscribers), Google AI Studio, Android Studio, and the Gemini Enterprise Agent Platform.
Sources: 9to5Google, DataNorth, Android Headlines