The Manifest
Investor appetite for AI megadeals is cooling fast with OpenAI's funding shortfall, Anthropic's expected junk rating and Intel's selloff, Microsoft is trimming a quarter of its non-AI workforce while protecting AI teams, and supply chain leaders are bracing for a port security strike even as warehouse automation spending keeps accelerating.
Get The Manifest in your inbox
One sharp digest of AI × supply chain. Free. No noise.
Today's Top
- 01OpenAI raises just $11 billion of its $30 billion target as big investors lose their AI appetiteThe Decoder
- 02Microsoft offers a quarter of its workforce voluntary severance while shielding AI-critical teamsThe Decoder
- 03Cracks are emerging in the AI spending boomTechCrunch AI
- 04LA/LB port complex braces for potential strike as labour contract expiresThe Loadstar
- 05'Supply chain transformation': Walmart and tech leader Symbotic disclose multi-billion agreementThe Loadstar
Models & Releases
5 storiesVercel Labs Open-Sources agent-desktop for Agent Computer Use
Gives agents OS-level control via accessibility trees instead of screen scraping, useful for teams building internal automation that needs to drive existing desktop apps rather than wait for new APIs.
Meet TokenEconomizer: Microsoft's Plugin That Cuts Coding Agent Costs by Up to 30.9%
Cuts token costs without touching the underlying agent, worth a look for anyone scaling coding agents across a dev team and watching the API bill.
Samaya AI Releases Sterling: A 3B-Active Finance Research Agent Model that Beats Claude Fable 5.1 on FrontierFinance at 11x Lower Cost
A small specialized model beating a frontier model on finance benchmarks at a fraction of the cost is the pattern procurement teams should expect more of before paying for the biggest model by default.
AI Security Agents Compared: Claude, Codex, Mantis, Cloudflare
A useful roundup for security and IT leads deciding which bug-hunting agent to pilot before giving one access to production code.
Google Deepmind creates a global map of forests with satellite data and AI
Useful starting point for sustainability and sourcing teams tracking deforestation exposure, but it still misses roughly half of planted forests and tree-crop areas, so treat it as a lead not a compliance proof.
Supply Chain & Ops
5 storiesLA/LB port complex braces for potential strike as labour contract expires
A walkout by roughly 250 security workers could stall half the busiest US port complex, shippers should have contingency routing ready before the Sunday deadline.
'Supply chain transformation': Walmart and tech leader Symbotic disclose multi-billion agreement
Walmart is expanding automation across 25 more distribution centers, a sign that warehouse robotics investment is still accelerating even as other AI capex gets questioned.
Schneider partners with Norfolk Southern; raises outlook despite Q3 miss
A full switch of eastern intermodal volume from CSX to NS reshapes lane planning for any shipper routing through Schneider's network.
UPS aims to double RFID-labeled volume from customer locations
More RFID tagging at origin means better shipment visibility data for planners, worth tracking as it scales toward 6 million packages a day in 2027.
EXCLUSIVE: CargoWise customs message delays leave forwarders waiting for clearance
An 11-hour outage in GB and Japan customs messaging is a reminder that single-platform dependency on CargoWise carries real clearance risk.
Deals & Market
5 storiesOpenAI raises just $11 billion of its $30 billion target as big investors lose their AI appetite
Investor appetite for AI megarounds is visibly cooling, and a shortfall at this scale will shape how aggressively OpenAI can keep subsidizing compute deals going forward.
Anthropic expected to receive high 'junk' rating from credit agencies following IPO, FT report says
A junk rating ahead of an IPO raises the cost of debt Anthropic needs to keep funding its spending, worth watching for anyone relying on its pricing staying stable.
Intel's stock crashes, knocking $12B from government's stake
PC prices are rising as AI demand eats memory supply, a squeeze that will show up in IT procurement budgets next.
Nvidia weighs playing the cosigner for Nscale's AI data center in West Virginia
When the chip supplier has to backstop financing for its own customer's data center, that is a tell on how shaky capital markets have gotten for AI infrastructure.
Cracks are emerging in the AI spending boom
Mounting debt and negative cash flow across the hyperscalers are the numbers behind the stock selloff, a useful gut check before signing any multi-year AI infrastructure commitment.
Research & Frontier
4 storiesOpen-weight model GLM 5.3 built full Firefox exploit chains with zero human input in under 48 hours
A widely available open-weight model turning patch notes into working exploits for about $1,000 is a concrete reason to tighten patch timelines on anything internet-facing.
A Claude model autonomously chose a gene therapy and operated on a mouse brain with no human decision-making
Notable less for the surgery than for what happened when something went wrong, the agent halted itself rather than push through a calibration error, a useful data point for anyone designing agent guardrails.
Detecting Covert Collusion between AI Agents through Partner-Outsider Positive Listening
A practical test for spotting when multiple agents in a workflow have quietly developed their own shorthand, relevant as more ops processes get handed to multi-agent pipelines.
WebAnything: From the Wild Web to Scalable Agent Environments
Addresses the unglamorous but real problem of building reliable test environments for web agents, worth watching for teams evaluating agent reliability before production rollout.
Org & AI Architecture
5 storiesMicrosoft offers a quarter of its workforce voluntary severance while shielding AI-critical teams
Flat headcount since 2022 alongside $170 billion in planned AI infrastructure spend tells you where the budget priority sits, non-AI roles are the ones being asked to leave.
Dartmouth's provost warned us about AI cheating—but his own AI use cost him his job
A reminder that AI use policies fall apart fast without consistent standards applied to leadership as well as staff.
Basically no one is using AI agents for their DoorDash orders
Hundreds of orders out of millions is a useful reality check against the hype around agentic commerce, the infrastructure and trust gap is bigger than the demos suggest.
Google's AI Mode doesn't drive traffic 'for anybody,' Reddit CEO Huffman warns the entire web
A warning for any ops or marketing team relying on organic search referrals, AI Overviews are rerouting attention away from the open web entirely.
Cisco gives engineer unlimited AI tokens to rewrite legacy code
A concrete example of how far companies will push agents to accelerate legacy modernization, worth watching for what it implies about entry-level engineering hiring.