The Manifest
Meta went all-in on AI agents at Connect while ocean carriers tightened contracted capacity and a murky chain-of-custody scare over F-35 parts exposed thin logistics visibility, and diesel export ban talk added fresh volatility to freight costs.
Get The Manifest in your inbox
One sharp digest of AI × supply chain. Free. No noise.
Today's Top
- 01Meta's AI agent Muse draws 500,000 users in a week along with claims it copied OpenClawThe Decoder
- 02Forwarder anger as Hapag-Lloyd cuts guaranteed-capacity allocationsThe Loadstar
- 03UPS or Ceva? Who lost the F-35 parts in Hong Kong?The Loadstar
- 04Export ban talk sinks diesel, boosts gasoline as White House eyes optionsFreightWaves
- 05Tesla wins lead role in 2,500-truck electric Class 8 orderFreightWaves
Models & Releases
5 storiesEverything new coming to Meta's AI agent Muse
Meta pushed hard on Muse at Connect, folding it into glasses and a new wearable, a sign agents are becoming the default interface rather than just another chatbot. Watch adoption numbers over the next quarter, not launch-day hype.
Meta introduces camera-free AI glasses
Lighter frames and 12-hour battery life matter more for anyone who'd actually wear these on a floor or a dock than the AI features do. Practicality, not novelty, decides whether wearables show up in real operations.
ChatGPT Voice gets closer to 'Her' with email, calendar, and Slack access
OpenAI is turning voice into an action layer that can touch your calendar and inbox, which is the real test of whether assistants save time or just add another app to babysit.
Gemini 3.8 text-to-speech
Google's new TTS models let you script tone and build two-voice dialogue straight from text, more useful for training material and IVR scripts than for consumer novelty.
Alibaba launches Qwen Audio 3.1 with new models and slashes AI audio prices by up to 95 percent
A 95 percent price cut on speech models is the kind of move that makes voice bots for dispatch and warehouse ops suddenly cheap enough to actually deploy at scale.
Supply Chain & Ops
5 storiesForwarder anger as Hapag-Lloyd cuts guaranteed-capacity allocations
A 40 percent cut to a supposedly guaranteed contract product is a reminder that guaranteed ocean capacity is only as good as the carrier's willingness to honor it when Asia gets disrupted. Buyers relying on these contracts should push for penalty clauses, not just promises.
Hapag tries to close Zim deal, Asia connectivity the crux
A Hapag-Zim tie-up would reshape Asia capacity just as carriers are already tightening allocations elsewhere. Worth tracking closely for anyone booking on those lanes into Q4.
UPS or Ceva? Who lost the F-35 parts in Hong Kong?
Nobody can even agree on which carrier handled the shipment, and that's the real story. Chain-of-custody failures on controlled cargo don't need malice to happen, they need bad visibility, and most networks have plenty of that.
Export ban talk sinks diesel, boosts gasoline as White House eyes options
A diesel export ban would ripple straight into carrier fuel surcharges and regional rate spreads. Lock in fuel assumptions carefully this quarter, this can move fast on policy headlines alone.
AI Saves Trucking Company $53,000 Annually
A small owner-operator used Claude to audit its own telematics data and found $53,000 in leaking labor costs. That is a better ROI story than most enterprise AI pilots, and it did not need a platform contract to get there.
Deals & Market
5 storiesEma raises $77M as AI starts eating into enterprise software and services
Fifty-plus enterprise customers including Google and Microsoft is a real signal that agent platforms are starting to displace seat-based software spend, not just bolt chatbots onto existing tools.
Enveda secures $311M to bring more nature-derived AI drugs into clinical trials
A $2 billion valuation for a biotech mining nature for drug candidates shows capital is still chasing AI-plus-life-sciences even as software funding gets pickier elsewhere.
Nvidia-backed Nscale keeps its biggest customer, Bytedance, out of its IPO filing
Omitting your largest customer from an IPO prospectus is a flag worth watching. Concentration risk in AI infrastructure deals is exactly what investors will dig into before pricing this one.
Inside Basecamp Research, the AI startup turning evolution into training data
A $140 million raise for a company building biological training data, backed by Nvidia and Anthropic's fund, shows where the next data moat fight is heading.
OpenAI is enlisting an influencer army to make it look 'good for the world'
Leaning on influencer sponsorships for a public image campaign is a tell that trust in AI labs is now a marketing problem, not just a policy one.
Research & Frontier
5 storiesAnthropic says its biology lab has already found something big
Anthropic is keeping humans in the loop on Claude's lab work for now, which is the more interesting detail than whatever the finding turns out to be. Autonomy in scientific discovery is still a choice, not a default.
XPRIZE Wildfire winners spotted fires within 10 min, but couldn't stop them
Detection is basically solved, suppression is not, which is the gap to expect anywhere you evaluate an AI monitoring pitch tied to physical risk.
Silent Failures in Agent-Tool Interaction: An Audit of ToolUniverse
A close look at where multi-tool agent pipelines quietly break, relevant reading for anyone stitching agents across ERP, WMS, and TMS systems.
Harness as a Language: A Minimalist Agent Framework With Maximal Expressivity
A cleaner way to build agent loops without bolting on extra engineering for every workflow, worth a look if your team is building internal agent tooling instead of buying it.
TwinCheck: Evidence-Grounded Negative-Twin Verification for Stateful Tool Agents
A verification method aimed at catching bad tool calls before they derail an agent's task, the kind of guardrail research that matters once agents are touching production systems.
Org & AI Architecture
4 storiesEven Americans who use AI every day are worried about it
Daily use isn't building trust, it's just familiarity. Leaders rolling out AI tools internally should expect skepticism to persist even as adoption climbs.
Anthropic engineer explains why Claude's writing got worse although the model got smarter
Optimizing models for other models to read has made them worse for humans to read, a useful reminder that benchmark gains and usable output are not the same thing.
Trump's China rivalry and 'AI race' delusion may endanger US, experts say
Framing AI policy as purely a race narrows the room for safety cooperation, which matters for anyone whose supply chain runs through both US and Chinese tech ecosystems.
OpenAI hires Patreon co-founder Sam Yam to lead a new Creator Product division
A new division built around creators rather than enterprises signals OpenAI is diversifying revenue lines beyond API and subscription seats, worth watching ahead of DevDay.