The Manifest
Ocean freight spot rates surged again on fresh July GRIs, Volvo put commercial driverless trucks on the Dallas-Houston lane, and enterprise AI buyers are beginning to throttle spending as costs scale alongside adoption.
Get The Manifest in your inbox
One sharp digest of AI × supply chain. Free. No noise.
Today's Top
- 01Spot rates surge again as carriers push through fresh July hikesThe Loadstar
- 02Volvo and AVI-SPL launch driverless freight operations on Texas corridorFreightWaves
- 03Companies rein in AI usage as costs strain budgetsft.com
- 04OpenAI tripled revenue to $5.7 billion in Q1 but burned through $3.7 billion to get thereThe Decoder
- 05Gulf land bridge gains momentum as DHL, Oman Air and GWC expand capacityThe Loadstar
Models & Releases
4 storiesData2Story turns a CSV file into a verified interactive news article using seven AI agents
A multi-agent pipeline from Oxford and Stanford that converts structured data into a cited, interactive article with verifiable sources shows that data-to-narrative workflows are production-ready. Supply chain teams that publish inventory, supplier, or logistics data internally should evaluate whether this kind of automation can replace manual data storytelling.
NVIDIA AI introduces SpatialClaw: a training-free agent that treats code as the action interface for spatial reasoning
Writing Python in a persistent kernel to compose 3D perception tools is a cheaper path to spatial reasoning than training a dedicated vision model. For warehousing and robotics teams, this reduces the compute barrier to deploying vision-based localization without custom fine-tuning.
ChatGPT keeps creeping toward becoming your AI personal assistant with new scheduled task controls
OpenAI's consolidated scheduled task page limits alerts to meaningful changes, which makes ChatGPT more viable as a lightweight monitoring tool for market conditions or supplier news. It is not a replacement for purpose-built supply chain monitoring, but it lowers the setup cost for ad hoc tracking.
Liquid AI introduces LFM2.5-Embedding-350M and LFM2.5-ColBERT-350M: dense bi-encoder and late-interaction models for fast multilingual search across 11 languages
Compact multilingual search models that run on edge devices reduce dependence on cloud APIs for searching supplier documents, contracts, or RFPs across languages. Teams with global procurement or multi-region supplier bases should benchmark these against their current vector search setup.
Supply Chain & Ops
5 storiesVolvo and AVI-SPL launch driverless freight operations on Texas corridor
Aurora-powered Volvo VNL trucks are now running commercial freight on the Dallas-Houston lane without a safety driver, making this one of the first live commercial autonomous deployments at scale. Watch load volumes and cost-per-mile disclosures over the next two quarters as the operational economics become public.
Spot rates surge again as carriers push through fresh July hikes
Shanghai-Rotterdam spot rates jumped 15% week-on-week after June 15 GRIs, with Asia-Europe routes up 12%. Shippers who waited on July contract negotiations hoping for relief should reopen those conversations now, not later in the month.
Gulf land bridge gains momentum as DHL, Oman Air and GWC expand capacity
DHL, Oman Air Cargo, and GWC are each adding capacity to the Muscat-Dubai corridor as Iran-related disruptions continue to make Red Sea routing unpredictable. Treat this as a structural route diversification, not a short-term workaround, when revising your carrier mix for the next planning cycle.
Highway looks to become the 'Plaid for Freight' as cargo theft goes direct
Highway is building a carrier identity verification layer for trucking at the load level, targeting cargo theft that now bypasses brokers entirely. For procurement and logistics teams, this shifts identity verification from a compliance checkbox to a direct risk-control input on load awards.
Bangladesh opens door to private air cargo operators in logistics push
Allowing private operator stations near international airports directly addresses chronic airfreight congestion for Bangladesh's garment export sector. Buyers sourcing from Bangladeshi suppliers should factor improved airfreight reliability into lead time assumptions over the next 12 months.
Deals & Market
4 storiesOpenAI tripled revenue to $5.7 billion in Q1 but burned through $3.7 billion to get there
Revenue tripled but so did losses, with stock compensation alone at $2.3B, and the $73B reserve does not guarantee pricing stability if a price war with Anthropic accelerates. Enterprise teams locking in multi-year AI contracts should treat current pricing as negotiable.
Billionaire Ambani wants AI in every call, app, and home
Reliance is embedding AI across telecom services used by more than 500 million people, setting a new baseline for what enterprise-facing digital tools look like in India. Operators sourcing from South Asian suppliers should note that AI-assisted communications are becoming infrastructure, not a differentiator.
Analysis: DSV CEO Jens Lund at a crossroads, next deal please
DSV has consistently grown earnings by integrating acquired forwarder networks faster than peers, and the market is watching what Lund targets next with Schenker still digesting. A new major acquisition would signal the ceiling for organic growth in asset-light freight forwarding and move the competitive baseline for everyone else.
The US says ASML's top chip tool may be in China, but how?
If an EUV tool reached China through gray-market channels despite export controls, the implications extend well beyond semiconductors. Any supply chain handling dual-use equipment components should audit end-user verification processes now, before regulators use this case to expand reporting requirements.
Research & Frontier
4 storiesHow to build a forecasting pipeline with TimeCopilot using foundation models and automated anomaly detection
This end-to-end tutorial covers rolling cross-validation across statistical, foundation, and GPU-accelerated models on panel data with anomaly detection included. Planners evaluating modern demand forecasting stacks can use this as a reference architecture before committing to a vendor.
New benchmark exposes how badly AI struggles with real knowledge work
The best current AI models fully solve only 3% of realistic knowledge work tasks on this benchmark. For supply chain teams evaluating AI for complex analyst work such as contract review or market analysis, this is a useful calibration before setting expectations with leadership.
OpenAI researchers show small doses of 'beneficial trait' training make AI models broadly safer and harder to manipulate
Reinforcement learning on specific traits like truthfulness improved safety across 44 of 53 benchmarks without domain-specific tuning. For teams building internal AI tools for planning or procurement, this suggests safety alignment is becoming a standard training layer, not a costly custom integration.
Deontic Policies for Runtime Governance of Agentic AI Systems
This paper proposes a runtime policy framework for AI agents that handles permissions and compliance constraints across organizational boundaries. Supply chain teams experimenting with multi-agent workflows for procurement or planning should read it as a preview of the governance layer they will need to build.
Org & AI Architecture
4 storiesCompanies rein in AI usage as costs strain budgets
Enterprises are throttling AI tool usage as budget pressure hits, and the correction is favoring teams that tied deployments to measurable cost or time outcomes from day one. For ops leaders, this is a signal to audit ROI on existing licenses before approving new tools.
Is AI ruining our skills? Early results are in, and they're not good
Nature's early findings suggest measurable skill degradation when workers consistently offload cognitive tasks to AI. Supply chain teams running AI-assisted planning should consider structured decision-review protocols to preserve analyst judgment, particularly for scenario planning and exception handling.
The US banned Anthropic's Fable 5 release, but the numbers don't seem to care
The US government pulled Anthropic's two newest models on national security grounds with limited public explanation. Enterprise buyers standardized on Claude should review vendor diversification options, as the pull demonstrates that model availability is now a policy risk alongside a commercial one.
Google appeals ruling that made it directly liable for AI-generated search overview content
A German court held Google directly liable for AI search summaries that falsely linked publishers to fraud schemes. For procurement teams using AI-generated supplier research or market intelligence, this ruling signals that liability for inaccurate AI outputs is shifting toward vendors, with implications for enterprise indemnification clauses.