The Manifest
Apple ships its Gemini-powered Siri while Big AI's push for a coordinated slowdown draws pushback from Nvidia, China and skeptics who call it a cartel move, and freight data shows rail consolidation, rate inflation and steady ocean demand reshaping network planning.
Get The Manifest in your inbox
One sharp digest of AI × supply chain. Free. No noise.
Today's Top
- 01Nvidia CEO Jensen Huang tells Trump 'we're not going to let [an AI slowdown] happen'TechCrunch AI
- 02Apple brings a fully revamped Siri built on Google's Gemini, but not to the EUThe Decoder
- 03Anthropic eyes Nasdaq listing as a second profitable quarter aims to win over investors ahead of a mega-IPOThe Decoder
- 04Container shipping health check: more room to manage the next supply waveThe Loadstar
- 05BNSF, CPKC, and CSX to request broad access to a combined UP-NS networkFreightWaves
Models & Releases
4 storiesApple brings a fully revamped Siri built on Google's Gemini, but not to the EU
Apple finally ships an assistant that can chain multi-step requests and read screen context, but early testers still hit hallucinations and gaps with personal data. Worth tracking for procurement teams evaluating device-side AI, though the EU holdout is a reminder that regulatory geography still shapes what ships where.
Apple releases iOS 27, macOS Golden Gate 27 with Siri AI and Liquid Glass refinements
The broader OS release matters more than the Siri headline for fleets running Apple hardware in warehouses and field ops. Note this is also the last macOS version supporting Rosetta, so plan Intel app retirements now.
AI bots 'Timmy,' 'Ren,' and 'Jackie' are flooding social media with slop
Autonomous agents posting on their own platform is a preview of what unmanaged agent sprawl looks like at scale. Anyone deploying agentic tools internally should be thinking now about guardrails before this becomes an operational headache rather than a curiosity.
Fashion app Daydream uses Apple Intelligence to help you shop the outfits in your camera roll
A small example of on-device AI turning saved images into shoppable product matches through Siri, no app open required. Retail and demand planning teams should watch this pattern spread into more commerce discovery flows.
Supply Chain & Ops
5 storiesContainer shipping health check: more room to manage the next supply wave
Teu-mile demand growth of 6.1% is outpacing vessel-supply growth of about 5%, with data-center and AI construction demand adding to cargo volumes. Carriers heading into autumn have more pricing power than the bear case assumed, so budget for firmer ocean rates through year end.
BNSF, CPKC, and CSX to request broad access to a combined UP-NS network
Rival railroads want guaranteed trackage rights before regulators bless a UP-NS merger, which is the real fight that determines whether shippers actually get competitive routing options afterward. Anyone with rail-dependent lanes should watch the STB filing closely, not just the merger headline.
Cass: TL rates jump 11% in August, freight shipments turn positive
Truckload linehaul rates up sharply year over year, and volumes finally inflected positive after 42 straight months of decline. If this holds, budget assumptions built around a soft freight market for 2027 need a second look.
Integrators pile into heavy freight as B2C loses its shine
DHL Express formally launched its up-to-3,000kg service in the US, following the broader integrator shift away from thin-margin parcel toward heavier, better-paying freight. Shippers with oversized or out-of-gauge cargo now have more integrator options than a year ago.
Shein opens automated distribution center Indiana
A 737,000-square-foot facility with goods-to-person automation shows fast fashion players investing in the same fulfillment tech traditional retailers have been rolling out for years. Worth benchmarking against your own DC automation roadmap.
Deals & Market
5 storiesAnthropic eyes Nasdaq listing as a second profitable quarter aims to win over investors ahead of a mega-IPO
The profitability claim leans on an adjusted metric that excludes stock-based compensation, so treat it as a marketing figure for now, not GAAP truth. Still, a mega-IPO filing would be a signal event for enterprise AI vendor pricing across the board.
OpenAI buys smartphone camera maker Glass Imaging for $300 million, report says
A hardware-adjacent acquisition suggests OpenAI is building toward its own device ambitions rather than just software. Watch whether this feeds into future consumer hardware plans that could compete with Apple's assistant push.
Superhuman acquires YC-backed notetaker Fathom as productivity platforms push for agentic work
Consolidation among meeting-notes and agentic productivity tools continues, with Fathom's 400,000+ monthly users now folded into Superhuman's roadmap. Buyers evaluating point tools in this category should expect more roll-ups like this.
Hub Group warns of Nasdaq delisting notice; flags H1 operating loss
An accounting error pushed Hub Group past its reporting deadline, triggering a delisting warning on top of a first-half operating loss. Shippers using Hub Group for intermodal capacity should ask about financial stability before locking in longer contracts.
OpenAI stuck fighting Musk antitrust suit after Apple finds a way out
Musk dropped Apple from his antitrust claims over ChatGPT integration but is keeping the pressure on OpenAI alone. Legal overhang like this rarely changes product roadmaps quickly, but it is a data point on how contested the assistant market has become.
Research & Frontier
4 storiesClay Mathematics Institute says the Navier-Stokes Millennium Prize Problem has 'apparently been settled'
A formal review is underway on a claimed solution to one of math's seven Millennium Prize problems. Not directly operational, but a reminder that AI-assisted proof work is accelerating progress on hard problems that eventually feed into better simulation and optimization tools.
Toward Self-Adaptive Physical AI: Can LLM Agents Manage Long-Horizon Physical Tasks?
This tackles the gap between LLM agents that talk well and agents that can reliably manage long-running physical processes like a warehouse or plant. Early-stage research, but the right benchmark to watch before trusting agentic control loops on the floor.
Converge Then Diversify: Decoupling Convergence and Diversity in Multi-Objective Bayesian Optimisation
A method for getting better Pareto-front solutions when optimizing multiple competing objectives with expensive black-box functions. Relevant groundwork for network design and sourcing tools that juggle cost, service and risk trade-offs simultaneously.
Generalized Agent Iteration: One Formal Framework for Iterative Policy Improvement and Recursive Self-Improvement
An attempt to formalize recursive self-improvement, a term that gets thrown around loosely in AI safety debates without a shared definition. Useful for anyone trying to separate genuine capability gains from marketing claims about self-improving agents.
Org & AI Architecture
5 storiesNot everyone is convinced that Big AI's proposed development slowdown is really about safety
Cohere's CEO calls the OpenAI-Anthropic-Google slowdown push a cartel by another name meant to lock out competitors, and the White House agrees it should not happen. Treat the safety framing skeptically until it is clear who benefits from slower frontier releases.
Microsoft's AI rulebook: readable thinking, no inner life, and definitely no rights
Microsoft's new code of conduct puts human control ahead of model autonomy and explicitly rejects any claim to machine consciousness, a direct contrast with Anthropic's more cautious stance. Governance teams drafting internal AI policies now have a second major vendor framework to benchmark against.
China fires back at U.S. AI safety warnings, calling them fearmongering to lock in American advantage
Beijing's Foreign Ministry and state media reject Amodei's slowdown calls outright, and China's security minister is pushing for faster AI infrastructure buildout instead. The safety-versus-competition split is now explicitly geopolitical, not just an industry debate.
OpenAI has hundreds of contract workers reading your ChatGPT conversations
Human reviewers are rating real conversations on a one-to-seven scale to curb flattery and human-like behavior, and the opt-out setting is buried and off by default. Any team piping sensitive prompts through ChatGPT should check that toggle today, not assume anonymization is enough.
AI leaders want to hit the brakes after years of reckless speed
The framing of safety-first slowdowns conveniently arrives right as frontier labs face rising compute costs and competitive pressure from open models. Read the sudden caution as much about market positioning as genuine risk management.