Key Highlights

  • Failing grades surged in UC Berkeley CS courses as instructors point to AI-driven academic dishonesty — 35.3% of CS 10 students and 10.6% of CS 61A students received F’s this spring, versus under 10% in prior years — a stark data point in the debate over how generative AI is reshaping (and eroding) foundational learning.
  • The U.K. forced Google to let publishers opt out of AI Search, with the CMA calling the Search Console toggle a “world first” — reframing yesterday’s Google “publisher controls” announcement as regulatory compliance rather than voluntary goodwill.
  • Alphabet raised a record-breaking $85B for Google’s AI business — an oversubscribed offering that topped Petrobras’s 2010 record — signaling investors’ near-insatiable appetite for AI infrastructure exposure.
  • NVIDIA shipped Nemotron 3 Ultra, a 550B-parameter (55B active) open MoE model purpose-built to orchestrate long-running agents at lower cost and reduced goal drift.
  • Theo and Sean Goedecke argue your prompts are tech debt — AGENTS.md/CLAUDE.md files decay silently with every model upgrade, making a January-tuned prompt actively harmful by February.

Analysis & Opinion

Failing Grades Soar With AI Usage, Dwindling Math Skills in Berkeley CS Classes — Daily Californian

The share of failing grades in several UC Berkeley computer science courses spiked far above historical norms in spring 2026, breaking the department’s own grading guidelines: 35.3% of CS 10 students and 10.6% of CS 61A students received F’s, against a guideline of roughly 7% D’s-and-F’s and a historical ceiling under 10%. Teaching professor Dan Garcia attributes the “primary driver” to a “vast increase in academic dishonesty” tied to students leaning on LLMs, compounded by weaker mathematical preparation and understaffing. The episode is a concrete signal of a tension educators have warned about: when AI can produce passing work, students may skip the struggle that builds durable skill, then fail when assessments demand genuine understanding. It also raises hard policy questions — whether to redesign assessment around in-person or oral exams, how to detect misuse fairly, and whether grading curves should even hold in an AI-saturated classroom. Expect this to become a recurring data point as more institutions report semester outcomes.

Publishers Will Be Able to Opt Out of AI Search, Thanks to New Regulation — TechCrunch

The U.K. has imposed legal guardrails on Google’s AI search, requiring the company to give publishers a way to opt out of being aggregated into generative AI features. Publishers can now flip a toggle in Google Search Console to keep their sites out of AI Overviews, AI Mode, and AI Overviews in Discover — recasting the “new controls for website owners” Google announced yesterday as regulatory compliance, not a voluntary concession. The Competition and Markets Authority calls putting publishers back in control of how their content is used a “world first,” and argues it strengthens news organizations’ hand in negotiating content deals with Google. The unresolved tension remains the same one publishers have flagged all along: opting out of AI features may also tank ordinary search visibility, so the “choice” could prove more theoretical than real. This is the clearest sign yet that the AI-search bargain between platforms and publishers will be settled by regulators, not goodwill.

Alphabet’s Record-Breaking $85B Raise for Google’s AI Business Is a Helluva Good Signal — TechCrunch

Alphabet’s stock sale was so oversubscribed that an initial $40B tranche raised $45B instead, with another $40B planned next quarter for $85B total — eclipsing Petrobras’s $70B record from 2010. The appetite reflects investors buying into a healthy parent (Alphabet posted $110B in Q1 revenue, up 22% YoY) rather than a debt-laden startup, making the raise a relatively low-risk vehicle for AI exposure. It’s a striking counterpoint to the week’s other narrative — enterprises struggling to prove AI ROI — underscoring that capital-markets enthusiasm and operational payoff remain on very different timelines.

New Products & Tools

NVIDIA Nemotron 3 Ultra Powers Faster, More Efficient Reasoning for Long-Running Agents — NVIDIA Developer

NVIDIA released Nemotron 3 Ultra, an open 550B-parameter Mixture-of-Experts model with 55B active parameters, built to orchestrate and plan in agentic systems where multi-turn token growth drives up cost and goal drift. The pitch is a model hierarchy: frontier reasoning models for orchestration paired with efficient models for high-volume execution and tool calling.

Ideogram and Reve Rethink How AI Images Get Made — The Rundown

Two image generators are moving past one-shot prompting toward layout- and layer-aware editing, letting creators manipulate composition and individual elements rather than re-rolling an entire generation. It’s a shift from “describe and pray” toward the kind of structured, editable control that professional design workflows actually need.

Lovable Signs Multiyear Deal With Google Cloud to Up Usage 5x — TechCrunch

Stockholm vibe-coding startup Lovable expanded its Google Cloud commitment roughly fivefold, gaining greater access to both Anthropic’s Claude (the dominant coding model) and Google’s Gemini. The deal sits downstream of Google’s $10B Anthropic investment and highlights how cloud providers are bundling rival frontier models to lock in fast-growing AI-native developers.

Meta’s AI Agent for WhatsApp Business Is Now Available Globally — TechCrunch

After nearly two years of testing in markets like India and Mexico, Meta’s renamed “Meta Business Agent” is now globally available in WhatsApp and Instagram DMs, able to answer questions, recommend products, book appointments, qualify leads, and escalate to a human — Meta’s bid to turn WhatsApp into workflow software for SMBs.

Coralogix Raises $200M on Bet That Someone Needs to Watch the AI Agents — TechCrunch

Software-monitoring firm Coralogix raised $200M (a $1.6B post-money valuation) just 11 months after a $115M round, betting that autonomous AI agents will need a dedicated observability layer for logs, metrics, and traces.

Google’s Dreambeans Will Turn Your Life Into a Cartoon — TechCrunch

Google Labs launched Dreambeans, an iOS/Android app that pulls data from Gmail, Calendar, Photos, YouTube, and Search history to generate AI-illustrated “stories” and lifestyle suggestions — a consumer experiment whose deep cross-service data access will draw privacy scrutiny.

Amazon Will Show AI Product Images When You Search — TechCrunch

Amazon will display AI-generated product images beneath search autocomplete to help shoppers visualize options when they lack the right search terms — a move critics flag as questionable on a platform selling real-world goods, where fabricated imagery risks misleading buyers.

These Two Founders Left Goldman and Meta to Build Voice AI for Overlooked Markets — TechCrunch

AethexAI raised $3M pre-seed to build low-latency voice AI for African and Middle Eastern markets, building its own small model and orchestration layer from scratch to handle localized dialects of English, French, and Arabic that mainstream tools weren’t designed for.

Bringing Secure Digital Identity and Payment Tools to More People — Google

Google announced expanded digital identity and payment features aimed at broadening access to secure verification and transactions.

Research

NVIDIA Research Unlocks Advanced Grasping, Smarter Autonomous Driving and Agent Training at Scale — NVIDIA Blog

Three CVPR 2026 papers share a common thesis — training at scale produces systems that generalize — spanning robot grippers that handle unseen tools, autonomous-driving models that reason fast enough on in-car hardware, and virtual agents exposed to many environments before deployment.

Interviews & Conversations

Your Prompts Are Tech Debt — Theo - t3.gg (20:29)

Reacting to a Sean Goedecke essay (transcript-based summary), Theo argues that prompt files — AGENTS.md, CLAUDE.md, skills, MCP servers, dynamically built system prompts — are a uniquely insidious form of technical debt because they decay silently: a prompt carefully tuned for one model version can become actively harmful after an upgrade, with no error to flag it. He notes that prompt engineering delivers real gains (Cursor’s harness reportedly lifts Opus quality 10–30% via system-prompt tuning and per-model “ungemini-ing” tweaks), but that this value is model-specific and perishable. His prescription is minimalism: pick a third-party tool (Claude Code, Codex, Cursor) and leave it as close to stock as possible to piggyback on vendors’ constant re-tuning; avoid unnecessary MCPs and skills; keep AGENTS.md to concrete facts, not behavior-steering fluff like “think step by step” or “I’ll tip you $200.” He points to a T3 repo whose AGENTS.md hasn’t been touched in two months despite multiple rewrites — and closes with Goedecke’s line: “Write your prompts yourself and delete them whenever you get a chance.”

Bill Ackman: Investment Strategy, What the Market Is Missing, How AI Breaks Businesses — All-In Podcast (29:59)

Pershing Square’s Bill Ackman (transcript-based summary) frames AI primarily as a disruption-risk problem: with near-unlimited access to compute, capital, and talent, “the probability of your being disrupted has gone up enormously,” making durability the hardest and most important thing to underwrite. He argues high-quality incumbents like Microsoft, Meta, and Amazon are being left behind as capital chases chips and energy — and that niche SaaS vendors charging monopoly-style prices (he names Salesforce as a worry) are most at risk, while broad low-per-seat platforms are safer. On enterprise adoption he’s blunt: he hasn’t “seen much success” beyond early legal/compliance back-office use, echoing the week’s enterprise-ROI skepticism and McKinsey’s stat that ~95% of enterprise AI initiatives fail. He treats frontier-AI bets (SpaceX/xAI, OpenAI, Anthropic) as late-stage venture investments underwritten on people-opportunity-context-deal, and praises OpenAI CFO Sarah Friar’s thinking on capital commitments while noting OpenAI’s spend-far-in-excess-of-revenue model is a “hard degree of difficulty.” Founder-led companies, he argues, have a structural edge in navigating AI disruption because founders have the authority and economic stake to make radical bets that hired CEOs won’t.

References

  1. Failing Grades Soar With AI Usage, Dwindling Math Skills in Berkeley CS Classes — Daily Californian, 2026-06-04 [blog]
  2. Publishers Will Be Able to Opt Out of AI Search, Thanks to New Regulation — TechCrunch, 2026-06-03 [blog]
  3. Alphabet’s Record-Breaking $85B Raise for Google’s AI Business Is a Helluva Good Signal — TechCrunch, 2026-06-03 [blog]
  4. NVIDIA Nemotron 3 Ultra Powers Faster, More Efficient Reasoning for Long-Running Agents — NVIDIA Developer, 2026-06-04 [blog]
  5. Ideogram and Reve Rethink How AI Images Get Made — The Rundown, 2026-06-04 [blog]
  6. Lovable Signs Multiyear Deal With Google Cloud to Up Usage 5x — TechCrunch, 2026-06-03 [blog]
  7. Meta’s AI Agent for WhatsApp Business Is Now Available Globally — TechCrunch, 2026-06-03 [blog]
  8. Coralogix Raises $200M on Bet That Someone Needs to Watch the AI Agents — TechCrunch, 2026-06-03 [blog]
  9. Google’s Dreambeans Will Turn Your Life Into a Cartoon — TechCrunch, 2026-06-03 [blog]
  10. Amazon Will Show AI Product Images When You Search — TechCrunch, 2026-06-03 [blog]
  11. These Two Founders Left Goldman and Meta to Build Voice AI for Overlooked Markets — TechCrunch, 2026-06-03 [blog]
  12. Bringing Secure Digital Identity and Payment Tools to More People — Google, 2026-06-04 [blog]
  13. NVIDIA Research Unlocks Advanced Grasping, Smarter Autonomous Driving and Agent Training at Scale — NVIDIA Blog, 2026-06-03 [blog]
  14. Your Prompts Are Tech Debt — Theo - t3.gg, 2026-06-03 [video]
  15. Bill Ackman: Investment Strategy, What the Market Is Missing, How AI Breaks Businesses — All-In Podcast, 2026-06-03 [video]