It already owns GitHub. Why not own the AI-era's GitHub, too? #
- Welcome to *. Cautious Optimism, a newsletter on tech, business, and power. Modestly upbeat
**Monday. **The United States and Canada are back in a trade war. A waste of economic growth and NATO unity. (Speaking of NATO, there’s a new push to bolster European competitiveness with tech roots, which is welcome.)
In better news, Intuit, Zoom, and nCino report earnings on Tuesday; Nvidia, CrowdStrike, Veeva, Okta, and Nutanix report Wednesday; and we’ll hear from Marvell, Autodesk, Workday, Affirm, Rubrik, SentinelOne, and BiliBili on Thursday.
Today, we’re talking about Instinct, Poolside, and why buying AI distribution is the new hot play. Let’s go! — Alex
📈Trending Up:Not being able to afford Fable…California-Paramount relations…Walmart in India…AI jitters…Robotaxi? …social media bans…post-Wildberries war…winning friends, influencing regulators…robot funding rounds…extreme weather…India-Russia relations…UK-Ukraine relations…Arctic drilling…bullets to fight back📉Trending Down:Ready cash…China’s moon mission…ready cash…the Hungarian deficit…freedom of the press…smuggling chips…private property in Russia
Quote of the Day: Data centers “basically dug their own grave for the problem that’s been caused for them, and that’s why they got the backlash they deserve,” says Texas Gov. Greg Abbott, referencing the speed at which data center construction has proceeded sans local buy-in.
It’s incredible how quickly a once-Lefty talking point has become the national mood. This crisis of confidence will take more than a few goodwill gestures to reverse, and I think that public opposition to the tech elite’s goals will engender even more anti-democratic agitation from the cohort.
Here’smore from the Texas GOP. Killer Instinct
We’re not even a year out from the dawn of the ‘Claw, but since the personal agent hit the market, we’ve seen the rise of its rival Hermes, the release of Anthropic’s Cowork app, and OpenAI’s Codex. Today, people can choose from a host of manicured and DIY AI agents.
Enter Instinct, a chat-media agent for individual use. The buzzy personal agent sports a Benchmark-esque corporate homepage, and has earned plaudits from venture capitalists, venture capitalists, and venture capitalists. So many tech investors received early access to Instinct — and tweeted about it! — that the product was criticized by *other *members of the industry for working too hard to gin-up investor interest.
How does a largely unreleased product generate enough discourse to earn a second-wave of commentary attacking the first? By apparently being pretty damn incredible.
Instinct manages to shine so brightly by taking some cool risks. All personal AI agents balance capability and safety. OpenClaw felt magical because it was light on guardrails and heavy on doing things for you. Instinct is similar on that front, and security concerns are already slowing its hype train, with users reporting that the system went rogue, got phished, and generally failed to properly disclose how it uses and retains data.
We’ll have to see what shape Instinct takes when it eventually launches broadly. But it seems clear that agents — be they the new Grok Bot or something from a startup — are still better at doing a range of tasks for users than they are at doing those tasks safely. That’s a massive opportunity for someone to fix. In the meantime, I’ll be in Codex.
A rare cybersecurity miss
Israeli cybersecurity startups have a pretty insane track record. Wiz was bought for $32 billion; CyberArk sold for $25 billion; Armis got $7.75 billion, and Cyera is now worth $12.0 billion. Snyk? $7.4 billion. The list goes on and on.
Hence my surprise when Minimus said it is shutting down. The secure image company’s founders previously built and sold cybersecurity startup Twistlock to Palo Alto Networks. Given their experience, capital, and knowledge of what it takes to win customers in the security space, you’d be forgiven for expecting the startup to go vertical like many of its peers.
But Minimus apparently went kaput because it “failed to build enough commercial momentum in an increasingly crowded market,” CTech writes.
The competition point is warranted. Cybersecurity startups raised $10.6 billion in the first half of 2026, more than how much the sector was able to raise in late 2022, 2023, and 2024. Since 2025, investor interest in security startups has re-accelerated. Not shocking, given that AI is shaking up how companies build, deploy, and consume software.
There’s more to secure than ever, before even taking into account Mythos and other hyper-capable AI models that may dissolve traditional digital security paradigms. Still, ample capital means lots of startups competing for the same deals, not to mention large public cybersecurity companies who don’t want to be disrupted, please and thank you.
- If I had to choose one non-AI startup thesis to bet on today, keeping the post-AI world secure would be it.
Why Nvidia bought Poolside
When the odd $6 billion + $1 billion Nvidia-Poolside deal was announced last week, we found ourselves scratching our heads. A non-exclusive license to the AI lab’s technology and offering 109 people jobs did not seem worth $6 billion.
At the time, ** CO **wrote that we failed “to understand precisely why Nvidia is spending so much money on Poolside’s technology,” unless the “company had a breakthrough” cooking.
We were *almost *right.
Subscribe to Cautious Optimism to unlock the rest.
Become a paying subscriber of Cautious Optimism to get access to this post and other subscriber-only content.