All Dispatches

Shopify is abandoning React Native for Swift and Kotlin, claiming AI coding agents have erased the cross‑platform savings that once justified the framework. The shift rewrites the economics of mobile development for large SaaS operators.

A new arXiv study shows that fine-tuning reshapes LLM internals but only a fraction of those changes actually drive task performance, raising questions about compute waste and vendor lock-in.

Google promises persistent, cross-device AI memory that stays private by keeping decryption keys on user devices. We break down the tech, the trade-offs, and what it means for enterprise AI budgets.

Black Forest Labs’ FLUX 3 Video promises multimodal generation but ships with vague pricing, proprietary lock‑in, and unproven enterprise safeguards that could leave IT budgets exposed.

Anthropic's August threat report frames AI misuse as novel when attackers simply repurpose safety tools for extortion and fraud. Here's why your SOC team is already behind.

Anthropic’s new terms cut off any company more than half‑owned by a Chinese parent, no matter where it’s incorporated. The move targets national‑security risks but could reshape SaaS procurement and supply‑chain due diligence.

OpenAI slashes Sol and Luna API prices by 50% while boasting Astra-level efficiency. But SaaS teams know the real cost hides in cache misses, effort levels, and promotional pricing ghosts.

OpenAI rolls out a fancier cache‑hit dashboard and new knobs for GPT‑6, but the underlying economics and lock‑in remain unchanged for enterprise SaaS workloads.

NVIDIA's new qualification program certifies power and cooling gear for its AI factory reference designs — if you build exactly their way, with their blessed vendors.

NVIDIA bets its Halos stack is the only way to deploy physical AI without regulatory annihilation. The vendors lining up suggest they're right.

Multiverse Computing frames transformer‑block deletion as an Ising spin glass, claiming a 23‑point MMLU boost at 50% depth. Here’s what the math actually promises—and where the hype evaporates.

V7 claims its Context Graph turns scattered enterprise files into agent memory, promising 21x faster deal screening and sub‑10‑hour insurance reviews. Here’s what the graph actually delivers—and where it could backfire.