The Manifest
OpenAI's Q1 revenue tripled to $5.7B while burn hit $3.7B, DHL moves autonomous vehicles into live Singapore hub operations, and AWS launches two services to give enterprise AI agents better business context and security controls.
Get The Manifest in your inbox
One sharp digest of AI × supply chain. Free. No noise.
Today's Top
- 01DHL transitions Zelostech autonomous vehicles to live Singapore hub opsFreightWaves
- 02OpenAI tripled revenue to $5.7 billion in Q1 but burned through $3.7 billion to get thereThe Decoder
- 03AWS says AI agents lack business context and security, launches two services to patch the gapsThe Decoder
- 04OpenAI's Codex can now watch you work once and repeat the task foreverThe Decoder
- 05Nobel laureate John Jumper is leaving DeepMind for rival AnthropicTechCrunch AI
Models & Releases
3 storiesAWS says AI agents lack business context and security, launches two services to patch the gaps
AWS Context builds a knowledge graph from internal corporate data so agents stop operating blind, while Continuum targets code vulnerabilities automatically. Both address the core enterprise complaint: agents that act fast but get things wrong because they lack internal context.
OpenAI's Codex can now watch you work once and repeat the task forever
Record a workflow once, Codex converts it into a reusable skill and runs it independently. For ops teams with repetitive digital tasks this is a credible path to automation without writing code. Note: EU, UK, and Switzerland rollout is currently blocked.
ChatGPT keeps creeping toward becoming your AI personal assistant with new scheduled task controls
OpenAI is moving ChatGPT toward persistent background monitoring, not just single-turn chat. Planners who track supplier news or freight rate shifts could set tasks that surface meaningful changes without manual checking each day.
Supply Chain & Ops
3 storiesDHL transitions Zelostech autonomous vehicles to live Singapore hub ops
Electric, driverless shuttles are now handling daily point-to-point transfers between logistics facilities at DHL's Singapore Advanced Regional Center. This is no longer a pilot. The pattern holds where routes are fixed and coordination between facilities is tight.
How long will this truckload market cycle last?
FreightWaves sees the current truckload tightness extending into 2027 or beyond. Shippers who expected rate relief this year should factor a prolonged cycle into mid-term procurement planning and contract negotiations now.
How to Build a Forecasting Pipeline with TimeCopilot Using Foundation Models and Automated Anomaly Detection
A practical walkthrough using foundation models for time-series forecasting with built-in anomaly flagging on real panel data. Supply chain planners can use this as a testbed to benchmark newer model approaches against existing statistical baselines before committing to a stack change.
Deals & Market
3 storiesOpenAI tripled revenue to $5.7 billion in Q1 but burned through $3.7 billion to get there
Revenue and burn both tripled year over year, with stock compensation alone exceeding $2.3B. The $73B reserve buys time, but a price war with Anthropic changes the math quickly. Enterprise buyers signing multi-year AI contracts should think hard about vendor stability in a sustained pricing squeeze.
Nobel laureate John Jumper is leaving DeepMind for rival Anthropic
The AlphaFold architect joins a wave of senior DeepMind departures to Anthropic. For supply chain teams evaluating long-term AI vendor relationships, the ongoing talent concentration at a handful of frontier labs is a consolidation signal worth tracking.
NYU finance professor Damodaran warns an AI crash could hit harder than the dot-com bust
Damodaran's argument is structural: AI's physical infrastructure is debt-financed in a way dotcom software was not, making a correction harder to absorb across the stack. Supply chain teams planning long-horizon AI vendor investments should price in this concentration and leverage risk.
Research & Frontier
2 storiesCisco AI Introduces FAPO: Pipeline-Aware Prompt Optimization With Step-Level Failure Attribution and Claude Code Orchestration
FAPO is open-source and targets a real problem: when a multi-step LLM pipeline fails, it is usually unclear which step caused it. FAPO attributes failures to specific steps and tests fixes systematically. Worth evaluating if your team runs chained AI workflows in planning or procurement.
Data2Story turns a CSV file into a verified interactive news article using seven AI agents
Oxford and Stanford researchers built a multi-agent pipeline that converts raw CSVs into interactive reports with source links for 93% of statements. The pattern maps directly to ops teams that need to automate routine data reporting without sacrificing auditability.
Org & AI Architecture
3 storiesSam Altman says a whole generation of researchers held AI back by underestimating what scaling could do
Altman's framing is self-serving, but the underlying pattern matters for enterprise AI strategy: teams that bet against scaling lost influence over how the technology developed. Organizations now setting AI roadmaps are making similar directional bets about which capabilities will materialize.
Signal's Meredith Whittaker wants you to remember that AI chatbots 'are not your friends'
Whittaker's point is about data collection, not sentiment. Any chatbot interaction is a data event logged by the provider. Procurement and ops teams should treat external AI tools the same way they treat any third-party SaaS with broad data access.
The EU doesn't really know what a deepfake is, and that's becoming a problem for retail
Retailers including Zalando, where 90% of marketing content is already AI-generated, are lobbying to narrow the EU AI Act's deepfake definition. The outcome shapes compliance obligations for any brand doing AI-generated visual marketing across European markets.