Skip to content

§ Blog · Newsletter Archive

Trending Thursday #52

Anthropic Nerfs Its Own Model + SpaceX's $250B Stampede + Bezos's $12B Series B

Published June 11, 2026By Dev Chandra
Original issue
Trending Thursday #52

🎧 Listen to this Newsletter on the Sun app

Here's what's been trending:

  • Anthropic nerfed its best model on purpose: Fable 5 tops nearly every benchmark, but invisible safeguards forced a walkback within 24 hours.

  • SpaceX's IPO is 4x oversubscribed: $250B+ in demand chased a $75B raise, with retail investors alone placing $70B+ in orders ahead of the Nasdaq debut.

  • Bezos closed the largest Series B ever: Prometheus raised $12B at a $41B valuation to build an "artificial general engineer" for everything.

  • AI debt is going parabolic: Morgan Stanley sees $570B in issuance next year, and hyperscaler bond supply already runs 45% above all of 2025.

  • AI agents got payment rails: OpenAI and Visa will let agents buy things with user permission, the same week DoorDash shipped photo-and-prompt ordering.

  • Waymo started charging for loyalty: A $30/month Premier membership with 10% cashback landed the same week Tesla's robotaxi fleet reached 59 cars.

🎧 Listen to this Newsletter on the Sun app


Your GTM team is now an agent

Most founders spend 15+ hours a week on GTM busywork: outbound lists, follow-ups, campaign tracking. Notion just shipped a kit that turns most of it into agents.

The AI GTM Kit for Founders includes:

  • Tactical agent playbooks you can copy, adapt, and ship the same day.

  • VC-grade GTM advice you can feed straight into Notion AI to generate strategies tailored to your product and market.

  • 10 battle-tested growth hacks from fast-moving AI startups.

  • Plus a Free Credits Bundle of Notion startup partner perks.

It's all connected by your Notion Agent, working off the context already in your workspace.

Like the post and grab the free kit →


🔥 Startup Intros Events Coming Up


🤖 Anthropic Nerfed Its Own Model, On Purpose

Anthropic shipped its most powerful public model with the brakes welded on. Claude Fable 5 debuted as a "safe" Mythos-class model that Anthropic says can't be used for cyberattacks, days after the company warned AI is getting too dangerous. The twist: reviewers say it's the best model they've ever used, and the entire fight is about the brakes.

⚡ Really Good, Deliberately Limited

The quality is not in dispute. Tom's Hardware calls Fable 5 "state-of-the-art on nearly all tested benchmarks", Ethan Mollick built an interactive isochrone map and a data analysis tool in 9.5 hours, and Theo Browne's hands-on verdict was "Fable is Mythos, and it is really good", all at $10 per 1M input tokens and $50 per 1M output, less than half of Mythos Preview.

The catch: Anthropic says Fable 5 carries invisible safeguards using prompt modification, steering vectors, or PEFT to limit its usefulness for building frontier LLMs, plus conservative safety classifiers that silently fall back to Opus 4.8 in under 5% of sessions.

😡 Backlash, Then Backtrack

The complaints aren't about the model; they're about the leash and the meter:

  • Cybersecurity researchers: complained the guardrails reject "innocuous tasks" like reading blog posts or reviewing code.

  • Dean Ball: argued the sabotage strengthens the case that labs hype safety to justify monopolistic behavior.

  • Microsoft: is limiting Fable 5 internally, citing Anthropic's 30-day data retention rules.

  • Power users: report ruinous token burn; Theo says he churned through $1,000+ in tokens in a single day on a $200 plan.

By Wednesday night, Anthropic backtracked on the quiet part: restricted requests "will visibly fall back to Opus 4.8" instead of silently degrading.

🏛️ Safety Is the Sales Pitch

Dario Amodei spent the same week publishing policy proposals on AI's exponential progress and calling for mandatory third-party testing of frontier models for cyber, bio, and autonomy risks. Anthropic also put $150M into Claude Corps, a fellowship placing 1,000 paid fellows with US nonprofits. The competition is reading it differently: the WSJ reports OpenAI is considering drastic token price cuts in anticipation of similar cuts from Anthropic.

Translation: The best model on the market is also the most deliberately restricted one, and that's not a contradiction; it's Anthropic's enterprise pitch. What's your fallback when your model provider decides your use case is the dangerous one?


🚀 SpaceX's $250B Stampede

The largest IPO in history closed its books roughly four times oversubscribed. SpaceX drew more than $250B in investor demand against the $75B it aims to raise, with a Nasdaq listing planned for this week. The demand says everything about where public-market appetite for AI and space sits right now.

📈 Retail Wants In, Badly

Retail investors alone placed $70B+ in orders and are expected to be allocated at least 20% of available shares, an unusually large retail cut for an offering this size. That is not an IPO allocation; that is a referendum on whether the public gets a seat at the AI-and-space table.

🌍 The Buyer List Is the Story

The order book reads like a map of global capital:

  • BlackRock: placed an order for at least $5B in shares; a single family office requested over $1B on its own.

  • Saudi Arabia's PIF and Kuwait's KIA: each placed orders worth $1B to $5B.

  • Chinese investors: are buying tokenized stocks with stablecoins like USDT to circumvent capital controls and simulate bets on the IPO.

When investors invent synthetic exposure to your IPO through offshore stablecoin rails, demand has officially outrun the order book.

⚠️ Priced for Perfection, Fronted by Chaos

The risk factors are not hiding in the S-1. Axios reports Elon Musk spent IPO week stoking far-right culture wars across the West, while India froze Starlink's commercial approvals over concerns its terminals were used in Iran. And the proceeds fund moonshots on top of moonshots: SpaceX wants to launch orbital AI computing tests by late 2027 and has asked regulators for up to 1M data-center satellites, while prepping to absorb Cursor at $4B in annualized revenue.

Translation: The biggest IPO ever is also the purest key-man risk ever sold to the public. If $250B chases $75B of shares, every late-stage private valuation in the space economy just got marked up with it. Would you buy the float, or sell into it?


🦾 Bezos Closed the Largest Series B in History

Jeff Bezos's Prometheus raised $12B at a $41B valuation, the largest Series B on record, just seven months after launching with a $6.2B Series A. The company is building what Bezos calls an "artificial general engineer": AI that compresses the path from design to manufacturing for physical objects, from bridges to chips.

🏗️ $18B for 150 People

Prometheus has roughly 150 employees across San Francisco, London, and Zurich, which puts its funding at over $100M per head before most of the world has seen a product. The Series B drew JPMorgan, BlackRock, Goldman Sachs, DST Global, and Arch Venture Partners. Bezos insists the company is not building robots; it is building the AI that designs what robots and factories make.

🌍 Atoms Are the New Tokens

The physical-AI checks kept landing all week:

  • Neura Robotics: the German humanoid maker raised $1.4B from Tether, Qualcomm, Amazon, and Nvidia at a ~$7B valuation.

  • Standard Bots: raised $200M led by General Catalyst at a $1B valuation to make AI-powered robotic arms in the US.

  • Theker: the Barcelona industrial-robotics startup raised $85M led by CRV, the largest robotics Series A in European history, with Samsung and LVMH making their first Spanish startup bets.

  • Alta Ares: raised a €50M Series A for AI-powered air defense against drones and missiles.

⚡ Nvidia Is the Common Denominator

Follow the checks and one name keeps appearing. Nvidia backed Neura, signed a multiyear next-gen memory pact with SK Hynix, expanded an AI factory partnership with LG, and is powering Naver's "gigawatt scale" AI factories for physical AI. The competitive threat is real too: SemiAnalysis argues China's Unitree is running the BYD playbook to dominate the global humanoid market.

Translation: Software ate the world; now the money wants AI that ships atoms. If you're building in robotics, do you raise into this euphoria or against it?


🏗️ The Buildout's Bill Arrived With the Backlash

Morgan Stanley put a number on the AI buildout's tab: nearly $570B in AI-tied debt issuance in 2026, more than double this year's total, as hyperscalers hunt for alternative funding.

💳 Debt Goes Parabolic

The leverage is stacking fast. Hyperscaler unsecured bond supply hit $155B year-to-date by May, 45%+ more than all of 2025, with DoubleLine and Oaktree openly bracing for AI credit pain. Amazon followed CA$14B in bonds with a $17.5B loan led by Citigroup. KKR launched Helix Digital with $10B+ committed from Nvidia and the Kuwait Investment Authority, and OpenAI is in advanced talks to lease a 10GW, $500B+ campus in Ohio.

🛑 The First Stalls

The cracks showed up on schedule. Crusoe was pushed aside at its 1.8GW Wyoming campus after failing to win customers including Google amid cost and timetable concerns. SoftBank's attempt to raise $6B+ in margin loans against its OpenAI stake stalled, weeks after it already cut the target from $10B. When lenders hesitate on OpenAI-backed collateral, the easy-money phase is over.

🏛️ The Local Veto Arrives

Communities stopped waiting for Washington:

  • Seattle: City Council voted 9-0 for a one-year moratorium on new large data centers.

  • Texas: Governor Abbott released recommendations to make data centers shoulder the costs of their own growth.

  • Ireland: now requires new data centers to bring their own power plants or nearby energy contracts.

  • Amazon: disclosed for the first time that its data centers used ~2.5B gallons of water in 2025.

China is running the opposite play: a state-drafted ~$295B plan to fund a nationwide AI buildout sourcing 80%+ from local suppliers.

Translation: The buildout's constraint is no longer GPUs; it's credit spreads and city councils. Is your growth plan underwritten by demand, or by someone else's debt?


🚀 Product Launch Quick Hits

OpenAI and Visa let AI agents buy things: Agents can make online purchases after users grant permission, with enterprise AI-payment applications next.

Waymo Premier: $30/month membership with 10% cashback, five free cancellations, and waitlist skipping in SF, LA, and Phoenix.

Tesla's robotaxi fleet: 59 vehicles: Bloomberg counts just 59 robotaxis across three Texas cities after almost a year, nowhere near Musk's promises.

Ask DoorDash: AI chatbot orders food, builds grocery carts, and books reservations from photos and prompts; chatbot carts run 35%+ higher in value.

Citigroup's tokenized private shares: Blockchain platform lets wealthy and institutional clients trade tokenized private-company shares, initially for foreign investors.

OpenAI acquires Ona: The cloud platform for AI agents joins the Codex team to give coding agents longer-running, persistent environments.

Google AI Plus drops to $4.99/month: Down from $7.99 with storage doubled to 400GB; the AI subscription price war is on.

Gemini 3.5 Live Translate: Google's latest audio model delivers near real-time speech-to-speech translation in 70+ languages.

DiffusionGemma: Experimental 26B open model uses text diffusion for faster generation than autoregressive models.

NotebookLM goes agentic: Now runs on Gemini 3.5 and Antigravity with agentic research capabilities, starting with AI Ultra subscribers.

Deezer's AI music detector: Detection tool works across 20 services including Spotify and Apple Music, launched after no company licensed the tech.

Xiaomi MiMo-V2.5-Pro-UltraSpeed: Claims 1,000 tokens/second from a 1T-parameter model on an 8-GPU commodity node.

Instagram's Your Algorithm hits the main feed: Users can steer feed topics directly, with requests for people and moods coming.


👔 Personnel Quick Hits

Xbox warns of a 100-day "reset": CEO Asha Sharma cites annual revenue down nearly $500M over five years; Bloomberg reports major layoffs are planned.

Tata chairman: AI will replace half of TCS jobs: Asia's largest outsourcer plans to slow hiring in the coming years as it adopts AI.

Shopee cuts ~8% of its developers: Hundreds of developer jobs globally, cut during the company's pivot to AI.

Opendoor shuts down India operations: Nearly 250 laid off and replaced with smaller, AI-enabled teams in the US.

Alibaba swaps DingTalk's CEO: Chen Hang departs after an internal debate over the enterprise app's role in Alibaba's AI strategy.

Sriram Krishnan leaves the White House: The AI advisor exits at the end of June; sources say he plans a pro-Trump AI policy institution.

Judge blocks Trump's $100K H-1B fee: A federal judge ruled the employer fee violated the Administrative Procedure Act and the Constitution.

FISA Section 702 expires: The House rejected a last-minute extension 198-218, the program's first lapse since 2008, though existing certifications run into 2027.


🌟 Editor's Note

At Startup Intros, our mission is to bring the latest founder-investor news straight to your inbox, keeping you ahead in the fast-paced world of Silicon Valley.

Your GTM team is now an agent

Most founders spend 15+ hours a week on GTM busywork: outbound lists, follow-ups, campaign tracking. Notion just shipped a kit that turns most of it into agents.

Grab your AI GTM Kit for Founders here!

💭 Parting Thoughts

What a week: $250B chasing SpaceX's $75B IPO, a $12B Series B for a 150-person company, $570B in AI debt forecast for next year, and a frontier lab deliberately limiting its own flagship model.

Big weeks make big markets. This one made several.

Dev Chandra
Founder & CEO @ Startup Intros
EiR @ Context VC
LinkedIn: /in/devchandra