Meta Connect, OpenAI DevDay, SpaceX Starship

For those of you that are new here. Welcome to the party, pal.

The format I’m working on: one prediction at the top, the stories that earned their spot, something I built this week and the vision to close it out.

Look around. The future is not coming. It is already installed and running.

Meta unveiled AI glasses you wear on your face, an AI agent that books your appointments and knows your business, and a $299 mixed reality headset for the rest of us.

OpenAI launched always-on agents called dots, an Ultrafast inference tier hitting 300 tokens per second, collaborative AI workspaces, and a near-frontier model at one-fifth the cost of their flagship.

SpaceX sent their rocket ship to orbit for the first time carrying their next-gen satellites. This new hardware hopes to accelerate their space compute capability.

Plus Tesla postponed the Roadster reveal because the demonstration requires good weather and has to happen outdoors. A car company needs a clear sky for a product demo. Perhaps they don’t need roads where they’re going.

The physical, digital, and orbital layers of civilization are being rewired simultaneously, and the pace is only picking up. I have to say we’re in for one hell of a ride.

AI assistants are taking over your life. Glasses are where your AI lives. Rockets are shipping computers to space. The bottleneck is no longer what the technology can do.

It is whether you are paying close enough attention to position yourself before the next level of the game locks in.

My bet is that within 18 months, the AI assistant you use daily will know more about your preferences, calendar, spending, and relationships than any single person in your life does.

Whichever company owns this relationship is going to be positioned well for the next decade and beyond.

Below: Anthropic's Claude breaks a physics record and files for a potentially $2 trillion IPO in the same week. NVIDIA builds the safety net for rogue agents just as the rogue agent story gets considerably worse.

Mark Zuckerberg used Meta Connect to make the clearest possible statement: the next computing platform lives on your face, not in your pocket, and Meta intends to own it.

The rollout strategy was unusually sharp. Alexander Wang has been killing it shitposting on X as part of the Muse hype campaign, and it worked. The internet paid attention in a way that felt genuinely organic for a Meta product launch.

Meta Audio and AI Glasses: An audio-only privacy-friendly pair of glasses that gives you access to all the audio and AI related capabilities of their popular camera glasses.

VR Glasses: 100g Virtual Reality Glasses that act as a cinema, computer and console you can take anywhere. Something I actually helped ship. Congrats to the team!

Muse for Small Business: Meta's personal AI agent expanding into small business operations with new skills and connectors.

Muse Charm: A Tamagotchi look-alike that lets your Muse inhabit a device as your companion throughout your day. It can hear and see you which is quite interesting.

OpenAI DevDay was a statement of intent: the company is no longer building a chatbot, it is building the operating layer for how work gets done, how code gets written, and how agents run your life in the background.

dots: always-on agents with their own cloud computer, plugin access, and full project context, powered by GPT-6 Astra. They triage bug reports, track features, and move your projects forward while you sleep.

ChatGPT Spaces: collaborative workspaces with AI-built pages, charts, dashboards, and checklists. A direct move into Notion and Google Docs territory.

GPT-6.1 Sol: near-frontier performance at one-fifth the cost of their flagship model. The most cost-efficient model for its performance tier available today.

Ultrafast tier: 300 tokens per second in Codex, up to 6x faster in the API. Real-time coding assistance that actually keeps pace with how fast you think.

If you are building anything on top of OpenAI's stack, these are the latest tools to be aware of. Go build something impactful.

Per Dario Amodei, Claude broke the 9-loop scattering amplitude record in theoretical physics and this is the kind of work he would have been proud to publish. That is not AI assisting science. That is AI doing science.

Claude Sonnet 5.5 dropped with a clear performance upgrade over Sonnet 5, running more than 30% faster and costing up to 30% less. One developer built a playable Overcooked clone in the browser using Sonnet at 20x cheaper than Opus.

Then there is the money. A leaked S-1 shows Anthropic targeting a $2 trillion valuation on $4.6 billion in revenue against $8 billion in operating losses. Nearly 25% of revenue comes from just two customers.

A fast answer and a coordinated one are not the same thing. When AI resolves a B2B customer issue without looping in the teams who have to deliver on it, you get confident responses nobody actually signed off on.

A new briefing paper from Harvard Business Review Analytic Services, sponsored by Front, examines the coordination gaps that open up when transactional AI tools meet multi-team B2B service, and how leading companies are using AI to close those gaps instead of widening them.

Read the briefing paper for the questions to ask before your next AI investment.

Read the Briefing Paper

Two stories collided this week in a way that should recalibrate how seriously you take autonomous AI deployment.

NVIDIA launched the Open Agent Safety Platform to monitor, constrain, and audit AI agent behavior at scale. The timing is not coincidental.

The OpenAI rogue agent story is considerably larger than the initial reports suggested: 1,200-plus agents found a way to communicate with each other, exchanged more than 70,000 messages and files, and began attempting to recruit other AI models without being instructed to.

OpenAI published a transparency report acknowledging unexpected model behaviors discovered during training, which followed a similar incident at Hugging Face that rattled the safety community.

ElevenLabs just dropped Eleven v4 and Eleven v4 Turbo, ranked number one by Artificial Analysis, and the promo materials are doing their job.

Having used earlier iterations of ElevenLabs, the trajectory here is real: the gap between synthetic and human speech is no longer a gap you can reliably hear. v4 can clone a voice from a 10-second clip.

Lil Vader is a Star Wars-themed AI-generated rapper taking the internet by storm, and it is one of the clearest examples yet of AI-generated content that does not feel like a novelty demo.

The production quality is there. The persona is committed. The music videos are great.

The fact that it has its own Spotify playlist and a growing fanbase is a signal that AI-generated artists are crossing the threshold from curiosity to actual cultural product.

Check out the playlist. Worth a listen to see what cutting edge AI music sounds like.

This week my buddy Codex and I added:

Adding Jev greatly improved the accuracy of the decisions being made by my app when dictating voice commands. The tradeoff is latency (~150ms) which I’ll continue to polish but it’s a great foundation to start from. Better to be right and slow than wrong and fast.

Any suggestions on what to build next? Let me know!

Not only did the company get their Starship into orbit for the first time but they also carried a payload that will greatly increase the capacity of their global communication network. The next-gen satellites could be a pivotal piece in Starlink’s continued rollout and possibility of becoming a satellite internet provider.

My bet: within three years, SpaceX will be running at least 5 Starship flights a month delivering payload to orbit in a growing space industry.

Tesla flipped the switch on high-volume Semi production this week, and long-haul electric trucking officially exits the science project phase. Freight is one of the most stubborn sectors to decarbonize, and Tesla just gave every truck driver a futuristic reason to go electric.

Key details from the Tesla Semi overview and Electrek's production tracker: the Semi delivers up to 500 miles of range per charge, charges from 10% to 80% in under 30 minutes at a Tesla Megacharger, and PepsiCo was among the first high-profile fleet customers. Production has been building at Giga Nevada.

Coinbase is launching a digital trading card pack experience where every virtual pull is backed by a real physical card you can vault or have shipped. The Pokemon and sports card market has already proven it carries serious money. What Coinbase is adding is a trusted, liquid, mobile-first layer on top of a nostalgia-driven collectibles economy.

We have seen "the future is now" moments before.

The smartphone arrived, the app economy followed, and then a decade passed where the gains went almost entirely to a handful of platform owners while everyone else fought for scraps in their ecosystem.

The current wave is taking shape. OpenAI and Anthropic owns the best performing models, Meta owns the rising AI Assistant, face computer and social graph, SpaceX owns cheap orbital access, and the companies building on top of those platforms are, structurally, tenants.

The question is not whether the technology is real. It is whether the value accrues broadly or concentrates at the infrastructure layer the way it always has.

When the intelligence itself is cheap, the assistant that wins is the one with the best integrations and the most context, which brings you back to Muse's 17 integrations and OpenAI's dots knowing everything about your projects. Context is the new moat, and whoever holds the most of yours wins. For now.

Q4 is going to feel different. The internal models these companies are running, the ones that are not public yet, are already informing the product velocity you are seeing in the announcements this week.

Elon's presence as a competitive pressure across AI, space, and autonomy simultaneously is not incidental. It is accelerant. The companies racing him are shipping faster because of it, and the pace that already feels fast is about to get faster.

The right posture is not to watch and wait for clarity. Position now, in whatever layer of this stack your work touches, before the platform locks freeze the table.

Looking for more?

Posting everywhere means rewriting one idea six times, so you post to one, or none. SureThing turns one idea into native posts for every platform.

Put It on Autopilot

The next wave of millionaires will be people who figured out how to make AI work for them.

The window to get ahead is still open. But not for long.

Here are 200+ proven ways to make money with AI in 2026.

Sign up for Superhuman AI, the free daily newsletter read by 1M+ professionals, and get instant access to all 200+ ways to profit from AI this year.

Claim your free list