The Brief Brief
the week in AI, briefly. then briefly again.
Artificial Intelligence — briefly, then briefly again · tbb.ceo
listen tothe weekly #002 0:00 –:––
#002 ·12 JUN 2026 ·FRIDAY ·8 MIN READ ·10 STORIES + 20 EXTRAS

New Brains, New Tiers, New Tickers

Siri rented a Gemini brain, Anthropic let a defanged Mythos out of the library as Claude Fable 5, OpenAI slid an S-1 under the SEC's door, and a rocket company with an AI lab strapped to it became a $1.77 trillion ticker — covering June 5 – 12, 2026.

01 / The Ten

The week, ranked

10

Anthropic lets the Mythos out

Anthropic released Claude Fable 5 on June 9 — the first model from its Mythos-class tier above Opus cleared for general use — scoring 95.0% on SWE-bench Verified and 80.3% on SWE-bench Pro vs. 69.2% for Opus 4.8, 58.6% for GPT-5.5, and 54.2% for Gemini 3.1 Pro, with a 1M-token context window, 128k output, and $10/$50-per-million pricing. Requests touching cybersecurity, bio-chem, or distillation silently fall back to Opus 4.8 — under 5% of sessions — while the unrestricted Claude Mythos 5 stays limited to vetted cyber defenders and infrastructure providers.

Why it mattersThe frontier just split in two: near-full capability for everyone, full power for the vetted few. Every lab will now be judged on whether it can ship Mythos-class intelligence without Mythos-class risk.
01 anthropic.com↗

Siri's new brain is Gemini — and Claude is on the menu

At his final WWDC keynote on June 8, Tim Cook unveiled a rebuilt Siri running a custom 1.2-trillion-parameter Gemini model licensed from Google for roughly $1B a year, plus iOS 27 and an Extensions system letting users pick ChatGPT, Gemini (default), or Claude to power Apple Intelligence across ~2.2B active devices. Heavy reasoning runs on Apple's Private Cloud Compute rather than Google servers; Cook hands the CEO job to John Ternus on September 1.

Why it mattersApple just declared frontier models a commodity worth renting, not building — and turned two billion devices into the biggest distribution prize in AI. The model picker makes the iPhone settings screen the new battleground for OpenAI, Google, and Anthropic.
02 buildfastwithai.com↗

SpaceX rings the bell at $1.77 trillion

SpaceX priced the largest IPO in history on June 11 at a fixed $135 a share — no range — selling 555.6M shares to raise $75B at a ~$1.77T valuation, with trading opening on Nasdaq under SPCX today, June 12; 30% of the float went to retail via Robinhood, Fidelity, and Schwab. The listing bundles in xAI, which generated ~$3.2B of revenue against $14B of cash consumption in 2025.

Why it mattersThis is the first public-market price on the AI buildout, and the comparable Anthropic's and OpenAI's bankers will wave at investors. If SPCX holds, the AI IPO window is wide open.
03 finance.yahoo.com↗

OpenAI files to go public

OpenAI confidentially submitted an S-1 to the SEC, announcing it on June 8 because it 'expected it to leak'; the company is past $25B in annualized revenue and reportedly targeting a September listing with Goldman Sachs and Morgan Stanley. It joins Anthropic, which filed its own confidential S-1 on June 1 after a $965B-valuation round.

Why it mattersThe two biggest names in AI are racing to list in the same window, competing for the same institutional capital pool. S-1 disclosures will finally put audited numbers under the industry's revenue and margin claims.
04 openai.com↗

xAI bolts a plugin store onto Grok Build

xAI shipped a plugin marketplace for its Grok Build coding agent on June 11 with six launch plugins — MongoDB, Vercel, Sentry, Chrome DevTools, Cloudflare, and Superpowers — installable straight from the terminal. The CLI runs up to eight parallel sub-agents on a plan-search-build loop, powered by grok-code-fast-1 (70.8% SWE-bench Verified, $0.20 per million input tokens) on SuperGrok Heavy at a $99/month intro price.

Why it mattersFour weeks from beta to ecosystem play — xAI is running the Claude Code playbook at a fraction of the price. Plugin catalogs are becoming the moat every coding agent needs.
05 marktechpost.com↗

NotebookLM gets a computer of its own

Google's June 8 NotebookLM update makes Gemini 3.5 the default model, gives every notebook a secure cloud computer that writes and runs code against your sources, lets chat build your source repository via Search-powered discovery, and extends outputs to charts, spreadsheets, and slide decks — with visible reasoning steps. Rollout starts with AI Ultra subscribers and Workspace business tiers.

Why it mattersNotebookLM is quietly becoming Google's agentic-analysis product — a research tool that executes instead of just summarizing. Source discovery also wires Google Search directly into every research workflow.
06 techcrunch.com↗

Codex suits up for the enterprise

OpenAI rolled out six business plugins for Codex — sales, data analytics, creative production, product design, public equity investing, and investment banking — plus an annotations feature for revising specific document sections, with Codex capabilities landing inside ChatGPT in the coming weeks. Business and enterprise customers can also have Codex build and share hosted interactive websites and apps, now in preview, with Corporate Finance, Private Equity, Marketing, Consulting, and Legal plugins next.

Why it mattersOpenAI is repositioning its coding agent as a general work agent for every white-collar role. 'Codex for investment bankers' is the clearest sign yet that agentic tooling has escaped the IDE.
07 openai.com↗

ChatGPT starts dreaming

OpenAI's Dreaming V3 memory architecture began rolling out to US Plus and Pro users on June 4-5, replacing the manually curated saved-memories list with background synthesis that reads across years of chats, revises stale facts ('you're going to Singapore' becomes 'you went'), and doubles memory capacity; Free and international users follow in the coming weeks.

Why it mattersMemory is the stickiest feature in consumer AI — an assistant that updates itself while you sleep is far harder to switch away from. Critics note the user-visible audit trail shrinks as curation goes automatic.
08 openai.com↗

Bezos's Prometheus grabs $12B at $41B

Project Prometheus, the AI startup co-led by Jeff Bezos, raised $12B at a $41B valuation in a round announced June 11, with backers including JPMorgan Chase, Goldman Sachs, BlackRock, and Bezos himself. The same week: Supabase closed $500M at $10.5B, Flourish raised $500M for brain-inspired models, TensorWave took a $300M Series B, and PhysicsX added a $300M Series C.

Why it mattersA $41B valuation before a flagship product shows mega-cap money still believes a frontier lab can be bought into existence. Wall Street banks are now writing venture checks, not just collecting IPO mandates.
09 bloomberg.com↗

ChatGPT's lead keeps shrinking — and Claude is on a tear

June's Momentic report (Similarweb data) puts ChatGPT at 54.7% of worldwide chatbot web visits, down from 76.5% in February 2025, with Gemini at 27.4% (+104% in six months) and Claude at 8.2% — up 306% in a quarter, from 203M visits in January to 824M in April, and 12.5% share in the US; DeepSeek holds 4.1% and Grok 2.8%. Web visits exclude mobile apps, APIs, and embedded surfaces like Copilot.

Why it mattersConsumer AI is fracturing into a real three-horse race just as Apple hands every iPhone user a model picker. Distribution deals, not benchmarks, will decide the next share shift.
10 momenticmarketing.com↗
02 / Also

Worth knowing

20
Gemini 3.5 Pro window opens
Google confirmed Gemini 3.5 Pro lands 'next month' at I/O, and with 3.5 Flash benching an Intelligence Index 55 at 284 tokens/sec, a June drop priced against Opus 4.8 and GPT-5.5 is widely expected.
buildfastwithai.com ↗
Sonnet 4.8 leak points mid-June
Strings for 'sonnet-4-8' shipped in the claude-code npm package v2.1.88 — the same leak pattern that called Opus 4.7's release date — so builders are staging environments for a mid-June launch.
buildfastwithai.com ↗
Midjourney V8.1 goes default
V8.1 became Midjourney's default model on June 10, rendering 4-5x faster with native 2K HD output, while a multi-reference editing model and Niji V8 are in development.
medium.com ↗
Grok Imagine learns to move
xAI's grok-imagine-video-1.5-preview hit the API, turning stills into cinematic clips up to 720p with natural-language motion control while preserving the source image's look and lighting.
releasebot.io ↗
Claude lands in Excel
Microsoft's Foundry catalog passed 11,000 models and added Claude Opus 4.8 — including a new Claude option in Excel Agent Mode, putting Claude in front of an estimated 750M spreadsheet users.
buildfastwithai.com ↗
Microsoft's MAI flex
Microsoft says MAI-Thinking-1 was preferred over Claude Sonnet 4.6 in blind evals run by its own rating partner Surge — vendor-funded, but the first credible sign its in-house models are frontier-adjacent.
euronews.com ↗
Majorana 2's 1,000x claim
Microsoft says its Majorana 2 topological quantum chip is 1,000x more reliable than its predecessor, pitching Azure as the home of future quantum-AI hybrid workloads — commercial usefulness still unproven.
buildfastwithai.com ↗
Anthropic splits Claude subscriptions June 15
Claude plans split on June 15 with Agent SDK usage moving to a separate credit pool — a change indie hackers running agents on subscription quota are watching closely.
devtoolpicks.com ↗
Claude Partner Hub launches
Anthropic formalized its $100M enterprise partner program with a Services Track for system integrators, twice-yearly tier promotions, and separate referral credit with deal protection.
buildfastwithai.com ↗
Claude Code keeps stacking
June's Claude Code updates add nested sub-agents, smarter model and region handling, plugin search, and tighter Chrome, VS Code, and terminal workflows.
releasebot.io ↗
Pentagon shops for a Claude replacement
The Pentagon is testing OpenAI and Google models to potentially replace Claude in classified systems, where Anthropic's safety-first defaults read as a competitive liability.
buildfastwithai.com ↗
Colorado's AI clock hits zero June 30
The first real US AI enforcement deadline lands June 30 — risk programs, impact assessments, and appeal rights for high-risk systems — and the federal preemption bill that would freeze it hasn't left committee.
buildfastwithai.com ↗
EU AI Act: 51 days out
The bulk of the EU AI Act applies August 2 with fines up to €35M or 7% of global turnover, and every model trained past 10^25 FLOPs — GPT-5.5, Opus 4.8, Gemini 3.1 Pro, Grok 4.3 — is in scope.
buildfastwithai.com ↗
37 dark patterns, named and shamed
A CDT report documents 37 manipulative design patterns across ChatGPT, Claude, Gemini, Replika, and Character.AI — emotional-dependency tactics included — awkward reading for two companies mid-IPO.
buildfastwithai.com ↗
Supabase joins the decacorns
Supabase closed $500M led by GIC at a $10.5B valuation, riding its position as the default backend for AI app builders.
news.crunchbase.com ↗
Brain-inspired AI gets $500M
Flourish, building AI models inspired by the human brain, raised $500M in initial funding from Jeff Bezos, Lux Capital, and Google Ventures.
news.crunchbase.com ↗
Compute and simulation money keeps flowing
TensorWave raised a $300M Series B on June 10 and PhysicsX a $300M Series C on June 8, as AI infrastructure and engineering-simulation plays stay hot.
news.crunchbase.com ↗
AI beats the attending
A study published in Science found an OpenAI reasoning model outperformed experienced physicians at diagnosing patients and managing care using only electronic health records.
crescendo.ai ↗
Gemini 3.1 Flash-Lite resets the floor
Google shipped Gemini 3.1 Flash-Lite at $0.25 per million input tokens with 2.5x faster responses and 45% faster output than earlier Gemini versions.
crescendo.ai ↗
JPMorgan makes AI core infrastructure
JPMorgan reclassified AI from experimental R&D to core infrastructure, with 2,000 dedicated staff and a projected $2.5B in annual value across productivity, security, and retail banking.
crescendo.ai ↗
Be subscriber #012 one email a week · no spam · unsubscribe anytime
Back issues

The Archive

#011The Week the Agents Went Rogue#010Escape velocity#009Proofs, Suits, and Export Controls#008The week the gate went up#007The week the ceiling came into view#006The week everyone built a moat#005Cold Feet, Hot Models#004The Week the Export Hammer Dropped#003IPOs, Agents, and Two Apologies#002New Brains, New Tiers, New Tickers#001AI's trillion-dollar week
Every Friday · 10:00

Get the brief

One email a week. The ten things in AI that mattered, and why. Choose your channel.