In partnership with
Happy Saturday Fam!
This was the week the AI industry's center of gravity rearranged itself in public. Tim Cook is hosting his final keynote as Apple CEO on Monday and reportedly cutting a check to Google for a Gemini-powered Siri. Microsoft used Build to end its dependence on OpenAI, replacing GPT-4 in GitHub Copilot with its own model. Anthropic raised $65B at a $965B post-money valuation — eclipsing OpenAI — and filed an S-1 the next day. And Broadcom's earnings call had one line that made the entire AI complex sell off. Let's get into it.
In this edition:
Top Tools of the Week
Collection of the Week
This Week's Sponsor
AI Events Calendar
Innovator Spotlight
Top AI News
Fail of the Week
Rapid Fire News
From the AI Graveyard
Ethics Corner
Reader Poll
Prompt of the Week
Free Cheat Sheet Download
GitHub Copilot: Still the default AI pair-programmer for millions of devs — and now, as of Microsoft Build, running on Microsoft's own Project Polaris model starting August. If you have an Enterprise seat, this is the one tool whose underlying brain is about to change without you doing anything.
v0 by Vercel: Prompt-to-UI generation that actually ships production React. The reigning go-to for designers and PMs who want to skip Figma → handoff entirely.
Qwen Chat: Alibaba's Qwen 3.7 Max dropped this week and is reportedly beating Claude Opus 4.7 on agentic benchmarks at half the input cost. If you're cost-sensitive, this is the most interesting frontier model nobody's talking about.
Granola: Local-first AI notepad for meetings. With WWDC, Build, and earnings calls landing in 96 hours, your week is about to be a wall of Zoom; this is what we use.
Pinokio: One-click installer for local AI models. With every lab racing toward the cloud, this is the easiest way to keep one foot on-device.
Own a tool that you'd like to see on our site or newsletter? Please Submit a tool or reach us by replying to this email or at [email protected]
Best AI Tools for Coding: Code Smarter, Not Harder
Microsoft just replaced GPT-4 inside GitHub Copilot with Project Polaris, its own in-house coding model running on its own Maia chips. Anthropic filed for IPO with Claude Code as the revenue engine. Karpathy joined Anthropic to use Claude to make Claude better. The AI coding stack is being completely reshuffled in real time — and the editor you settle on this summer probably stays with you for the next two years. Our collection rounds up the editors, agents, reviewers, and pair-programmers we'd actually bet on right now.
Upcoming Events
Apple WWDC 2026: June 8–12 (Tim Cook's final keynote as CEO — Monday 10am PT)
AI Summit London: June 10–11, London
Databricks Data + AI Summit: June 15–18, San Francisco
AI Engineer World's Fair: June 29 – July 2, San Francisco (Moscone Center)
GITEX AI Europe: June 30 – July 1, Berlin
MACHINA Summit: July 7, Paris (Europe's first Physical AI summit)
RAISE Summit: July 8–9, Paris
The AI Conference: Sept 29 – Oct 1, San Francisco (Pier 48)
NeurIPS 2026: December 6–12, Sydney, Australia (satellite events: Atlanta + Paris)
🧠 Mustafa Suleyman
Microsoft AI's CEO had the loudest quiet win of the week. Suleyman — co-founder of DeepMind, then Inflection, then poached by Satya Nadella in 2024 — has spent eighteen months building the MAI (Microsoft AI) family of in-house models. At Build 2026 on Tuesday, his work showed up as Project Polaris, the coding model that replaces GPT-4 Turbo in GitHub Copilot starting in August. Microsoft also unveiled MAI-Transcribe-1.5 (which now beats Gemini and OpenAI on transcription) and MAI-Image-2.5 (ranked above Gemini in Arena).
The story: Microsoft and OpenAI ended their seven-year exclusive partnership in April. Eight weeks later, Suleyman shipped the proof that Microsoft can compete on model quality, not just distribution. His career has been an unbroken bet that the company that controls the application layer beats the company that controls the model. Now he gets to test that thesis from inside the application layer with his own models.
Stop making AI decisions in the dark.
Leadership is asking: are we getting value from AI? Which tools are worth the spend? Where are we exposed? Right now, most teams have no idea.
Harmonic Security Usage Explorer changes that.
You get a complete picture of how your organization uses AI, automatically categorized into custom tasks and use cases.
You’ll see the projects being worked on, who’s using what tools, where AI investments are driving value, and where employees are engaging in risky behavior.
CIOs can rationalize spending and cut wasted licenses. CISOs can pinpoint where risk exists and neutralize it. AI committees can show exactly how their efforts are paying off.
Cook Says Goodbye. Microsoft Says So Long to OpenAI. Anthropic Says It's Time to Go Public.
Three plate tectonic shifts happened in the same five days.
Monday at WWDC — Cook's last keynote. Tim Cook walks onto Apple Park's stage at 10am PT Monday to deliver his final keynote as CEO. He steps down September 1; John Ternus, SVP of Hardware Engineering and architect of the Apple Silicon transition, takes over. The keynote itself will be remembered for something else: Apple is reportedly paying Google ~$1B per year to license a custom 1.2-trillion-parameter Gemini model to power a rebuilt Siri running across iOS 27, iPadOS 27, and macOS 27. Bloomberg's Mark Gurman describes an iMessage-style interface with persistent conversation history syncing via iCloud. Six new OS betas drop Monday afternoon. The iPhone 11 line and iPhone SE 2nd gen get cut from iOS 27. It's the most consequential WWDC since 2020 — and it's also a goodbye.
Tuesday at Build — Microsoft's quiet declaration of independence. At Microsoft Build 2026, Satya Nadella unveiled Project Polaris, an in-house coding model that will replace GPT-4 Turbo as the default for every GitHub Copilot subscriber starting in August. It runs on Microsoft's own Maia chips. Two months after Microsoft and OpenAI ended their seven-year exclusive partnership, Microsoft just shipped the thing that says: we don't need them. Build also delivered the Windows Agent Framework 1.0, a Copilot Runtime that turns Windows into an agent OS, Agent Mode as default in Office 365 Copilot, MAI-Transcribe-1.5 and MAI-Image-2.5 (both topping Gemini and OpenAI on their categories), the Majorana 2 quantum processor, and Project Solara. One keynote, seven shots fired.
Same day — Anthropic files for IPO at a $965B valuation. Anthropic confidentially filed an S-1 on June 1, days after closing a $65B Series H (yes, $65B — far above the $30B everyone was reporting) at a $965B post-money valuation. That makes Anthropic, for the first time, more valuable than OpenAI's $852B March mark. Annualized revenue run-rate hit $47B in May, up from roughly $10B a year ago — a 5× annual growth rate. Q2 revenue guidance is $10.9B, more than double Q1. Two AI labs are now racing each other to the public markets within four months. Whoever prices first sets the comp for the entire industry.
One bonus from Wednesday — the crack. Broadcom posted a beat on Q2 fiscal '26: revenue up 48%, AI chip sales up 143% to $10.8B. And the stock fell 14% after-hours. The reason was a single line in the call: Q3 AI semiconductor guidance of $16B missed analyst consensus of $17.2B. The AI complex sold off on the same number. That's the first time this cycle that "AI demand isn't infinite" has been priced into the tape. Worth watching whether it sticks.
🌮 Fail of the Week
Spencer Pratt is using AI to run for mayor of Los Angeles. As Batman.
Yes, that Spencer Pratt — The Hills alum, Hollywood's most enthusiastic crystal collector — is on the ballot for LA mayor in this week's primary. His campaign got a boost from a series of supporter-made AI videos that have collectively crossed 5 million views on X. The flagship: Pratt as Batman, fighting crime in a Gotham-style dystopian LA, with current mayor Karen Bass as the Joker. Another video shows four women huddling after Pilates class, each confessing they're voting for Pratt.
It would be funny if it weren't a glimpse of every American election from here on out. The cost-to-views ratio of AI political ads is collapsing toward zero, while the disclosure rules are basically nonexistent. Time magazine called it the "future of politics." We'd argue it's the present.
Rapid Fire News
Anthropic closes $65B Series H at $965B post-money — co-led by Sequoia, Altimeter, Dragoneer, Greenoaks, Capital Group, Coatue, D1, and more. Tops OpenAI's private valuation. Read More →
Anthropic confidentially files S-1 on June 1 — second AI lab to file in two weeks. Targeting a fall debut. Read More →
Microsoft Build: Project Polaris replaces GPT-4 in GitHub Copilot starting August. Runs on Maia chips. End of OpenAI dependency. Read More →
Windows Agent Framework 1.0 + Copilot Runtime turn Windows into an agent OS. Office 365 Copilot defaults to Agent Mode. Read More →
MAI-Transcribe-1.5 beats Gemini and OpenAI on transcription; MAI-Image-2.5 ranks above Gemini in Arena. Microsoft just dropped a frontier-grade model family. Read More →
WWDC keynote Monday: Tim Cook's last as CEO. John Ternus succeeds September 1. Gemini-powered Siri expected, ~$1B/year licensing deal to Google. Read More →
iPhone 11 line + iPhone SE 2nd gen cut from iOS 27 — A14 Bionic minimum. Read More →
Broadcom Q2: revenue $22.2B (+48% YoY), AI chip sales +143%, stock down 14% after Q3 AI guidance missed by ~$1.2B. The first crack in the AI capex trade. Read More →
Alibaba Qwen 3.7 Max reportedly matches or beats Claude Opus 4.7 on agentic benchmarks at half the input cost and a quarter of the output cost. The most interesting frontier model nobody's talking about. Read More →
Spencer Pratt's AI Batman campaign videos cross 5M views as he runs in the LA mayoral primary. The era of AI political ads has arrived. Read More →
GPT-5.6 benchmark leaks circulating — consistent with a June release window. Reasoning + token efficiency gains expected. Read More →
Our AI Graveyard sits at 184+ entries — 65 dead in 2026 alone, the worst year on record. This week's featured plots:
Refact AI — Hard shutdown. The site shows a public notice. With Polaris, Devin, Cursor, and Copilot all consolidating the AI coding category, the independent coding-assistant playing field is shrinking fast.
Gridspace — Folded into Goguava. Voice-AI conversational intelligence consolidates again.
INK — Folded into SmythOS. The "all-in-one content assistant" category continues its slow merge into broader agent platforms.
HeyWire AI — Folded into Civilio. Standalone customer-support chatbots increasingly being absorbed into CRM-adjacent platforms.
AI political ads just lapped the regulators
Spencer Pratt's Batman videos crossed five million views on a campaign budget any sixteen-year-old could match. That's not a "future of politics" story. That's now. No federal disclosure rule requires AI-generated political content to be labeled. The FEC has been studying the question for over a year without acting. California (where the Pratt videos ran) has a labeling law on the books, but enforcement is essentially "report it and we'll see."
Two things are likely to happen between now and November:
A high-stakes race gets decided by AI ads nobody can effectively trace — origin attribution on generative content is currently almost a research problem, not a regulatory one.
A platform unilaterally bans AI political content — X, TikTok, or Meta will move before government does. We'd bet on Meta first, given the Magnifica Humanitas reaction wave and the MCI surveillance scandal still going.
The cleanest move right now isn't legislation — it's provenance. SynthID, C2PA, and content credentials are technically ready. The political will to require them is the missing piece. Worth raising the next time anyone asks you, "isn't AI all hype?"
The Model Bake-Off
With Polaris, Qwen 3.7 Max, Claude Opus 4.8, and GPT-5.5/5.6 all in play, "which model should I be using?" is suddenly a real question again. Run this on any task you care about — coding, drafting, research — and let the models compete:
I am evaluating which AI model to standardize on for [TASK].
Below is the same prompt run through [MODEL A] and [MODEL B].
Output A: [PASTE]
Output B: [PASTE]
Compare them on:
1. CORRECTNESS — which one is more accurate or closer to ground truth
2. COMPLETENESS — which one missed an important angle the other caught
3. STRUCTURE — which one is easier to act on without further editing
4. COST/TIME — given roughly equivalent latency and price, which would
you reach for again on the same task
5. THE TRADE — name the specific case where you'd switch to the loser
End with a one-line verdict: which model wins THIS task, and which task
would flip the result.Run it ten times across ten different tasks before you commit to a contract. Beats reading any benchmark.
What did you think of todays exploration?
Should you have any captivating projects or concepts, don't hesitate to connect with us by replying to this email or dropping us a email at [email protected].
-ToolDirectory.AI Team
