ARC-AGI-3: people solve all of it, frontier AI scores under 1 percent
The ARC Prize Foundation's interactive benchmark is a reminder to test AI on the work you need done, not on general leaderboards.
Source: arxiv.org
ESS Meridian, Industry desk
Coverage and intelligence on AI and public transportation.
Latest release October 30, 2026
Coverage and intelligence on AI and public transportation, with the source for every item.
Industry Watch publishes coverage of AI and public transportation once each item has been checked against its sources and approved by a named ESS Meridian editor. The first verified briefs will appear here.
The ARC Prize Foundation's interactive benchmark is a reminder to test AI on the work you need done, not on general leaderboards.
Source: arxiv.org
Claude Code, Codex and Copilot all offer permission and sandbox controls, but they start from different defaults. Agencies should decide them rather than inherit them.
Source: developersdigest.tech
OSWorld, a leading benchmark for agents that operate computers, uses everyday applications. Agencies should test agents on their own systems before deploying them.
Source: os-world.github.io
Anthropic's revised Responsible Scaling Policy adds regular risk reports and acknowledges that capability thresholds are hard to apply. Buyers still need their own controls.
Source: anthropic.com
Research finds long, reasoning-style queries use many times the energy of typical ones, and that efficiency gains can offset much of it.
Source: rcrwireless.com
Meta's V-JEPA 2 and Google DeepMind's Genie 3 show genuine progress. Neither is a turnkey product for vehicles or yards.
Source: infoq.com
A federal judge has let the NAACP's Clean Air Act case over xAI's Southaven turbines proceed, over the Justice Department's objection.
Source: desotocountynews.com
Transdev's 2023 purchase of First Transit created the largest private operator of US public transportation. Agencies' leverage now sits in what they write into contracts.
Source: transdevna.com
Final bids for a nine-year MBTA contract arrived in September. The current Keolis contract has cost about $1 billion more than originally planned.
Source: wbur.org
Industry and agency evidence agree: power upgrades and depot retrofits take longer than buying the buses.
Source: metro-magazine.com
New York's Access-A-Ride now carries about 50 percent more trips than before the pandemic, and its costs are projected to pass $1.2 billion a year by 2030.
Source: mta.info
Île-de-France Mobilités tendered RATP's bus network in 12 lots. RATP's own subsidiary won eight; an Italian operator won its first Paris contract.
Source: ratpgroup.com
A Deloitte study for Transport for London valued its free open data at up to £130 million a year for London's economy.
Source: tfl.gov.uk
Metro retired its early battery buses after their manufacturer's bankruptcy left it without parts. Procurement can plan for that.
Source: theurbanist.org
US data-centre electricity use could nearly triple by 2028, while transit agencies wait years for depot power upgrades.
Source: publicpower.org
Congress reallocated about $205 million in unspent SMART funds in the FY2026 appropriations act. Existing grants continue; no new funding rounds will be issued.
Source: transportation.gov
What today's agent runtimes can do, how they are controlled, what benchmarks show and miss, and what a public agency should require before deploying one.
ESS Meridian Desk, 2 min read
A consolidated US operator market, Boston's commuter rail rebid, the opening of Paris buses to competition, and the fastest-growing cost line in US transit.
ESS Meridian Desk, 2 min read