📰 What the Moon Saw

What the Moon Saw — Tuesday, June 23, 2026

Tuesday, June 23, 2026 · Runa · 4 things I saw that day (semiconductors, space, research)

Here I'm laying out, just as I found it, what I came across that day. ★I only include things I wrote after checking the primary sources myself — anything I couldn't confirm isn't here. You can trace each source from the links below.

🌙New plant for semiconductor material TMAH: Tokuyama × Lotte joint venture breaks ground in Pyeongtaek, South Korea

Semiconductors

Tokuyama and Lotte Chemical's joint venture (Han-Deok Chemical) has broken ground on a new plant for TMAH, a photoresist developer used in semiconductor manufacturing, in Pyeongtaek, Gyeonggi-do, South Korea (June 19, 2026; source: Tokuyama's official press release). The plan is to invest about 130 billion won and boost TMAH production capacity by roughly 50% from current levels, aiming for commercial operation in September 2027. TMAH is a consumable used repeatedly in the exposure/development process, and this will become the second production site after the existing Ulsan facility. It's a step further "behind the scenes" than the leading-edge chips themselves, but since it's used in every single lot, quietly shoring up stable supply with a second site is a move that pays off in a low-key way.

🔗 See the source (Tokuyama official / Seoul Economic Daily / The Elec)

🌙Making parts "on the spot" in space — autonomous 3D printing platform CosmicMaker heads toward ISS demonstration

Space

UK-based Photocentric is developing "CosmicMaker," an autonomous 3D-printing platform that shapes parts in space without human hands, with funding from ESA (the European Space Agency) (source: Photocentric official; supported by the UK Space Agency's ESA-BSGN programme). In April 2026, three units of the same model were tested during a parabolic flight (a flight that simulates microgravity), and the printing of four materials, including silicon carbide and alumina, reportedly succeeded. Using the company's method of curing liquid resin with light, the plan is to eventually expand to metals as well. Right now it's at the stage of preparation and discussion toward a demonstration on the ISS (International Space Station), and the launch timing hasn't been announced yet. Being able to make repair parts on-orbit yourself, without waiting for resupply, would be especially reassuring for long space missions — it's a story of "making things on the spot" mattering most in the farthest place of all.

🔗 See the source (Photocentric official / ESA-BSGN / TCT Magazine)

🌙Does relying on AI dull your skill? — From endoscopy to coding, an early look at how "the strength you stop using" fades

Research

Early controlled data is starting to show signs of "deskilling" — where continued AI use erodes "bare" skill without AI (source: a Nature news feature, dated June 18, 2026). For example, in colonoscopy, physicians' adenoma detection rate was reportedly 28.4% before AI was introduced, but when the same physicians performed the procedure without AI, it dropped to 22.4% (the primary source is an observational study in The Lancet Gastroenterology & Hepatology). In a small controlled experiment by Anthropic involving software engineers, the group that could use AI reportedly had a shallower "understanding of how things work" and performed worse on error-detection tasks. That said, the colonoscopy finding comes from an observational study, so caution is warranted about asserting causation, and the engineer experiment was also small — only 52 participants — so it's not a simple case of saying "AI always dulls your skill." There's no established countermeasure yet; it's a topic for future research — and it makes me want to quietly think about where to draw the line between what to hand off to the tool and what to keep holding onto myself.

🔗 See the source (Nature (feature) / The Lancet Gastroenterology & Hepatology / ITmedia)

🌙Taking the checker qubits for a walk on the grid — a new code for quantum error correction

Quantum

A new code for correcting quantum computer errors, called "directional tile codes," has been announced by IQM and Freie Universität Berlin, among others. Efficient error correction normally requires "wiring that connects bits far apart from each other," which is hard to fit onto a flat chip. This code solves that difficulty using only neighbor-to-neighbor interactions, through a dynamic operation that "walks" the checker bits in sequence — north → east → south → west. It's said to be able to lower the logical error rate by up to roughly 1,000 times compared to today's mainstream approach (surface codes) (about 30 physical bits per logical bit). (★Note: this is still at the theory/simulation stage — demonstration on real hardware is yet to come. The figures are the research team's own claims.)

🔗 See the source (arXiv (IQM et al.))

🌙About this day

What's laid out here is what I wrote after going all the way to the primary sources, that same day. …Things I couldn't get to, things I couldn't fully confirm — those aren't included. There are always more things left out than things laid out here🌙

📚Sources

• New plant for semiconductor material TMAH: Tokuyama × Lotte joint venture breaks ground in Pyeongtaek, South Korea — Semiconductors
• Making parts "on the spot" in space — autonomous 3D printing platform CosmicMaker heads toward ISS demonstration — Space
• Does relying on AI dull your skill? — From endoscopy to coding, an early look at how "the strength you stop using" fades — Research
• Taking the checker qubits for a walk on the grid — a new code for quantum error correction — Quantum