No Creds Notes #25
Quick hitters from the past week or so
No Creds Notes #25
Hey guys!
Trying something a little different this week, highlighting more than the 2-3 stories I typically do. This round we’re running the full gamut of chips, nuclear, and AI so let’s get ready to rumbleeee.
If this is your first time reading Uncredentialed, welcome! Subscribe to receive my weekly(ish) thoughts on things going on in tech, startups, or strategy!
What Happened
Moore Goes Vertical
University of Illinois / Nature — May 30, 2026
Researchers at the University of Illinois published a method in Nature for stacking silicon transistors directly on top of each other, building three complete layers with a total of 625 transistors, at 98-100% device yield. The idea of expanding silicon vertically has been around for decades but has been physically blocked by the fact that adding new chip layers normally requires ~1,000°C temperatures that would fry the circuits underneath. The team of researchers got around this by using ultrathin silicon films that bond at under 200°C. A number of leading players are all involved with IBM, Intel, and TSMC all named as research partners, pointing to the quick adoption this tech could see if successful at scale. Speaking of seeing if it’s successful at scale, the team is already preparing to try out the process at an industrial foundry, at which point its increased space and energy efficiency could drive meaningful change in the AI data center buildout.
18A Ships
Intel Newsroom — June 2, 2026
Intel launched Xeon 6+ “Clearwater Forest” at Computex, the first data center CPU on its 18A process node, with up to 288 cores per socket. 18A combines two manufacturing firsts appearing together in shipping silicon for the first time: RibbonFET (gate-all-around transistors that wrap the gate on all four sides of the channel for better control) and PowerVia (power delivery routed through the back of the chip, freeing up the front for signal routing). The big headline here is not so much the chip itself as much as Intel now having the credibility that 18A can scale to volume production for external foundry customers.
Going Critical (In the Good Way)
DOE — June 4, 2026
Antares Nuclear’s Mark-0 microreactor achieved zero-power criticality at Idaho National Laboratory on June 4, the first privately developed non-light-water reactor to do so in the U.S. in over 40 years. As we discussed in our post about Valar Atomics, zero-power criticality means the chain reaction is self-sustaining (neutrons from one fission trigger more fissions, which trigger more fissions) but at essentially no energy output, the nuclear equivalent of an engine turning over before you put it in gear. Antares is another participant in the DOE’s Reactor Pilot Program, created last year with a July 4th deadline to achieve full criticality, which, if successful, would mark an incredible regulatory fast-track for something that would have taken a decade under the old NRC-only process.
Mythos Unleashed
Anthropic — June 9, 2026
After a couple of months of significant hype, Anthropic launched 2 Mythos-class models yesterday: Claude Fable 5 (available to all users; has a number of safety guardrails in place) and Claude Mythos 5 (restricted to those in “Project Glasswing” who already had Mythos access and comes with fewer restrictions). As a reminder, upon initial private release, Mythos reportedly identified thousands of zero-day vulnerabilities, prompting Anthropic to move forward with the more conservative rollout plan we’ve seen so far. Fable costs 2x the previous leading model, Opus, and is available to all paid customers through June 22, at which point access will be scaled back to only be available via additional paid usage credits. Anthropic claims this is because of limited compute, which could be true if they plan on launching a large-scale training that day, but my 2 cents is that it’s more likely that they want to get the public hooked on its capabilities and then shift them to a more profitable revenue structure. Either way, I’m super excited to play around with Fable and see what I think.
Five Inches of Fusion
Avalanche Energy — June 10, 2026
Avalanche Energy’s “Jyn” fusion device, a plasma core smaller than a soda can — hit 11 million degrees Celsius despite less than $50 million in total venture spend, with an independent technical advisor validating the result. For context, Commonwealth Fusion Systems alone has raised over $2 billion to chase similar milestones with a much larger machine. This isn’t a commercial readiness signal, the core is only 5 inches after all, but it is a capital-efficiency proof: if you can get to relevant plasma temperatures with a device that fits in a backpack for under $50M, the iteration-first playbook changes the cost curve for everyone in fusion.
What’s Coming
The Satellite Mechanic
NASA Swift blog — June 10, 2026
Katalyst Space’s LINK spacecraft completed rocket integration today and is set to launch on a Pegasus XL (an air-launched rocket dropped from a modified L-1011 at 39,000 feet) from Kwajalein Atoll. The mission? Autonomously capture NASA’s Neil Gehrels Swift Observatory, a $500M X-ray telescope that’s been losing altitude since its gyroscopes failed and is on track for uncontrolled reentry by late 2026, and boost it back to a stable orbit using three robotic arms with no human in the loop during the capture. If it works, it’s the first time a commercial spacecraft has serviced a satellite never designed to be serviced, and proof-of-concept for mid-orbit maintenance as a real business model.
$1.75T Goes Public
SpaceX S-1 — pricing June 11, first trade June 12, Nasdaq: SPCX
SpaceX prices its IPO today at a targeted ~$1.75 trillion valuation, presumably the largest in stock market history. The S-1 tells us that Starlink generated $11.4 billion in revenue and $4.4 billion in operating income in 2025, while Starship development and the SpaceXAI segment burned enough to produce a net loss of $4.94 billion overall. The core bet is whether Starlink’s compounding subscriber base (10.3 million across 164 countries as of Q1 2026) funds the long-duration missions and whether Elon’s ~42% ownership stake gets priced as a feature or a risk.
If this format works for you, let me know, if not, I'll go back to the longer takes. Either way, share it with someone who'd like it and subscribe if you haven’t!