the update! - Today's Update

★ Daily Update

Wednesday, July 1, 2026

Anthropic's Claude export ban keeps reshaping the model landscape, with Sakana's new Fugu model hitting 93.2 on LiveCodeBench and multiple Asian labs shipping Mythos-like alternatives to fill the gap. DeepSeek open-sourced DSpark, claiming up to an 85% inference speedup, while Cognition's Devin Fusion harness cuts coding costs 35% by routing between frontier and cheap models - both signs that token economics, not raw capability, are now the lever teams are pulling. Enterprise AI spend keeps climbing regardless: Big Tech's data center commitments have topped $850B and RBC's latest CIO survey shows budgets still growing past pilot mode, even as reporting suggests AI coding tools are speeding up code without speeding up actual delivery. Meanwhile Okta rolled out agent lifecycle governance for regulated industries, and a repo-jacking attempt against Anthropic's own plugin ecosystem is a fresh reminder that agent tooling inherits classic open-source supply-chain risk.

Frontier Models & the Export-Ban Fallout

Sakana's Fugu launches with 93.2 LiveCodeBench score after Claude ban [Xtra!]

A non-US lab closes the coding-benchmark gap just as Anthropic's export ban reshapes who gets frontier coding models.

More: Sakana Fugu Launches With 93.2 LiveCodeBench Score After Claude Ban

Filed under: ai, coding-agents, benchmarks

Asian AI startups launch Mythos-like models [Xtra!]

Anthropic's export ban is spurring a wave of regional competitors built to Mythos's spec.

More: Asian AI startups launch Mythos-like models

Filed under: ai, geopolitics, models

DeepSeek open sources DSpark [Xtra!]

An 85% inference speedup released as open source could reset cost expectations across the industry.

More: DeepSeek open sources DSpark, a new framework to speed up LLM inference by up to 85%

Filed under: llm-inference, open-source, systems

Devin Fusion mixes frontier and cheap models to cut costs 35% [Xtra!]

A template for cost-aware agent design as teams move from flat subscriptions to per-token billing.

More: Devin Fusion

Filed under: ai-agents, coding-agents, cost-optimization

Enterprise AI Economics & Governance

Big Tech's AI data center commitments balloon past $850B [Xtra!]

Capex commitments of this scale lock in years of AI infrastructure spend regardless of near-term ROI questions.

More: Big Tech's AI data center commitments balloon past $850B

Filed under: infrastructure, data-centers, industry

Enterprise AI budgets keep growing despite cost concerns [Xtra!]

CIOs are past pilot mode - AI budgets keep expanding even as per-token cost anxiety rises.

More: Enterprise AI budgets are still growing, despite cost concerns

Filed under: enterprise-ai, industry, budgets

AI coding tools are speeding up code, not delivery [Xtra!]

Faster code generation isn't translating into faster shipped features - a warning against vanity metrics.

More: AI coding tools are speeding up code, not delivery

Filed under: engineering-productivity, ai-coding, delivery

Okta brings AI agent governance to regulated environments [Xtra!]

Agent identity and lifecycle governance is becoming a first-class enterprise requirement, not an afterthought.

More: Okta is the first independent and neutral identity platform to bring AI agent governance to highly regulated environments

Filed under: ai-agents, security, enterprise

Research & Open-Source Releases

RoadmapBench tests agents across real version upgrades [Xtra!]

A harder, more realistic bar than single-shot code tasks for evaluating coding agents.

More: RoadmapBench: Evaluating Long-Horizon Agentic Software Development Across Version Upgrades

Filed under: benchmarks, ai-agents, research

DiScoFormer unifies density and score modeling in one transformer [Xtra!]

An architectural simplification relevant to diffusion and generative modeling pipelines.

More: DiScoFormer: One transformer for density and score, across distributions

Filed under: research, generative-models, transformers

RL Beyond the Verifiable [Xtra!]

A sharp look at extending RL techniques to tasks without clean, verifiable reward signals - a core bottleneck for agent training.

More: RL Beyond the Verifiable

Filed under: reinforcement-learning, research, ai-agents

Google Cloud proposes an Open Knowledge Format for AI data sharing [Xtra!]

A proposed open standard for AI-ready data sharing could reduce vendor lock-in for structured knowledge feeding into models.

More: Google Cloud proposes Open Knowledge Format for AI-ready data sharing

Filed under: data, open-standards, ai

Product Launches, Culture & Supply-Chain Security

Gemini's personalized AI image generation now free for US users [Xtra!]

Google removes the paywall on personalized image gen, raising pressure on competing products.

More: Gemini's personalized AI image generation is now free for US users

Filed under: product-launch, image-generation, google

Google Cloud will sell specialist AI models built for science [Xtra!]

Cloud providers are productizing vertical, domain-specific science models as a distinct line, not just general LLMs.

More: Google Cloud will sell specialist AI models built for science

Filed under: product-launch, science, google-cloud

Salesforce employees confused by company promoting Anthropic's Claude in Slack [Xtra!]

Internal tooling choices expose awkward tensions between corporate AI partnerships and competitive positioning.

More: Salesforce employees are confused about why the company is promoting a competitor inside Slack

Filed under: industry, culture, ai-vendors

Repo-jacking attempt against Anthropic's Claude Community Plugins [Xtra!]

A supply-chain vulnerability in community plugin repos shows how fragile trust chains are for agent tooling.

More: Repo-jacking Anthropic's Claude Community Plugins (And the SHAs That Saved Them)

Filed under: security, supply-chain, ai-agents

Ablo: a collaboration layer for AI agents [Xtra!]

A new open-source project targets the missing coordination layer for multi-agent systems.

More: Ablo - The collaboration layer for AI agents

Filed under: ai-agents, open-source, infrastructure