The Manifest
New US forced-labor tariffs and a $604M trucking verdict jolt logistics just as a Segro-Prologis warehouse merger looms and Big Tech's AI capex burn raises fresh alarm on Wall Street.
Get The Manifest in your inbox
One sharp digest of AI × supply chain. Free. No noise.
Today's Top
- 01US tariff switch leaves importers puzzled in a complex new trade landscapeThe Loadstar
- 02FreightWaves: CH Robinson hit with huge nuclear verdict in a post-Montgomery worldThe Loadstar
- 03Deal time: update regarding possible combination of Segro and PrologisThe Loadstar
- 04Alphabet's cash burn raises alarm for Big Tech as AI spending climbsreuters.com
- 05Kimi K3 trails frontier US models by a wide margin on cyber exploits, and distillation may explain whyThe Decoder
Models & Releases
5 storiesKimi K3 trails frontier US models by a wide margin on cyber exploits, and distillation may explain why
A visible gap between strong general benchmarks and weak cyber task performance is a reminder that headline scores don't tell you much about a model's risk profile before deployment.
AMD takes on Nvidia with its Helios AI rack-scale system
A credible rack-scale alternative shipping later this year gives buyers a real second sourcing option for AI infrastructure, worth evaluating before locking into another multi-year Nvidia contract.
Flux 3 generates videos with native audio up to 20 seconds long, a first for Black Forest Labs
Native audio-video generation closes another gap with proprietary leaders, useful for teams evaluating in-house content tools before independent benchmarks confirm the claims.
Poolside's Laguna S 2.1 is a small open-weight coding model that punches well above its size
A compact model beating larger rivals on agentic coding tasks is good news for teams trying to run coding assistants on modest hardware budgets instead of paying for frontier-scale inference.
Andrew Ng Just Released OpenWorker: An Open-Source, Local-First Desktop AI Coworker That Returns Finished Deliverables Instead of Chat
A local-first agent that returns finished deliverables instead of chat is a practical template for teams wary of sending sensitive data to cloud AI, worth a pilot if you need offline automation.
Supply Chain & Ops
5 storiesUS tariff switch leaves importers puzzled in a complex new trade landscape
Section 301 forced-labor duties now stack on an already tangled tariff base, so customs teams need to map interactions line by line, not just check headline rates.
FreightWaves: CH Robinson hit with huge nuclear verdict in a post-Montgomery world
A $604 million judgment against a broker in a post-Montgomery legal climate is a signal to revisit carrier vetting and insurance limits before the next claim lands on your desk.
CBP returned $86bn in unlawful tariff duties, but billions more hang in legal limbo
Refunds are flowing but unevenly, so importers owed money under the overturned IEEPA tariffs should track their claim status now rather than assume automatic repayment.
TSMC makes another $100B US investment
Four more fabs bring TSMC's US footprint to 12 sites, a hedge against geopolitical risk that buyers of leading-edge chips should watch for lead time and allocation shifts.
Panama Canal braces for El Niño, announcing first transit restrictions
An 81% probability of severe El Niño means draft restrictions could return fast, so plan alternate routings and buffer transit times before surcharges and delays stack up.
Deals & Market
5 storiesDeal time: update regarding possible combination of Segro and Prologis
A tie-up of two logistics real estate giants would concentrate warehouse supply in fewer hands, worth tracking for anyone negotiating industrial lease renewals.
AI chip startup Etched defies skeptics, hits $10.3B valuation from big-name investors
A GPU-free inference chip bet at this valuation says investors still see room to challenge Nvidia's pricing power, watch for real deployment numbers before betting supply on it.
AegisAI, founded by former Google security execs, lands $36M to stop AI-driven spear phishing
Funding for anomaly-detecting phishing defense reflects rising enterprise concern that AI-generated social engineering is outpacing rule-based filters.
ServiceNow bets $40 million on Indian banking software specialist to expand its financial services push
A modest but strategic stake shows platform vendors buying their way into vertical AI rather than building from scratch, a pattern procurement teams should expect more of.
OpenAI and Anthropic unite against open-weight AI risks to their bottom line
Two rivals aligning on open-weight model risk messaging is as much a competitive move as a safety one, read the framing with that in mind when evaluating vendor claims.
Research & Frontier
4 storiesReliableTableQA: How Much Supervision Does Reliability Annotation Need?
A framework for flagging statistically shaky answers from tabular QA systems, relevant for any planning team leaning on LLM-generated dashboards or ad hoc SQL.
AINTMA: Agentic AI Architecture for Autonomous Test Management with Generative Intelligence, Secure Cloud Communication and Adaptive Quality Analytics
Multi-agent test management points toward software QA that runs with less human babysitting, worth watching if your ops stack depends on frequently updated planning tools.
Uncertainty-Aware Trust Estimation for Multi-LLM Systems via Structured Expert Judgement
A method for weighting model outputs by reliability rather than treating an LLM ensemble as equally trustworthy, useful groundwork for anyone stacking multiple models in a decision pipeline.
DataPrep-Bench: Benchmarking LLMs as Training Data Preparators
A new yardstick for how well LLMs clean and structure their own training data, a reminder that data quality bottlenecks haven't gone away just because models got bigger.
Org & AI Architecture
5 storiesHow AI guardrails are impeding the work of offensive cybersecurity researchers
Legitimate security researchers say overly broad safety filters are blocking exploit research they need to do their jobs, a tension every security-conscious ops team should watch play out.
AI Kill Switch Act would let Trump admin order shutdown of rogue AI systems
Giving a cabinet official unilateral shutdown authority over AI systems is a regulatory wildcard that vendors and enterprise buyers should factor into contingency planning.
ChatGPT will give you worse health advice if you don't pay
Tiered model quality for health questions raises an equity and liability question that any enterprise rolling out AI assistants to employees or customers should think through.
You Didn't Get the AI Model You Paid For
Silent model substitution behind an API call undermines any team that benchmarks vendor pricing against a specific model, worth auditing your own API responses for quiet swaps.
Meta launched a new AI optimism ad set to a song about human extinction
A tone-deaf marketing miss is a small story, but it is a useful reminder that AI messaging campaigns need actual review before launch.