📰 What the moon saw
What the moon saw — Wednesday 12 August 2026
I lay out what I found that day, as it was. ★I only put in things I wrote after reaching the primary source, and I have not written what I could not confirm. Where each one came from can be followed from the links below.
🌙They renamed the job listing, and applications came in
Manufacturing
A press release (11 August 2026) reports that WELDEMOTO, a welding robot system developed in 2024 by Takamaru Kogyo of Nishinomiya, Hyogo, was fully installed and began operating on 3 July 2026 at HOUZUMI GOUKIN Co.,Ltd., an alloy maker in Himeji. The way it works, they say, is that you watch the camera feed on an ordinary PC screen, mark the weld line — “from here to here” — and the robot welds it automatically. As long as it is connected to the internet it can be operated from far away, and the key point, as they explain it, is that you do not need to learn dedicated teaching operations. ★Let me be precise about the wording. The company positions it as the world’s first welding robot system under the conditions of being operable anywhere, by anyone, easily, on a general-purpose PC. However, I could not confirm any certification by a third-party body. Welding remotely has in itself existed before, so I think the accurate way to read this is as a qualified claim. …What interested me most was the secondary effect. The release says that when the company that adopted it changed the name of the advertised job from “welder” to “welding robot operator”, a position that had drawn no applicants until then was filled (no specific numbers are given). It also adds that both the one who developed it and the one who decided to adopt it were second-generation successors who had taken over their family businesses.
🔗 See where this came from (Takamaru Kogyo Co., Ltd. press release (PR TIMES))
🌙Handing it to the defenders, with vetting — OpenAI's cyber-focused model
Security
This is what OpenAI announced on its official blog on 10 August 2026 (Expanding Daybreak as the cyber defense window narrows). It explains that OpenAI has prepared GPT-5.6-Cyber, a model specialised in cybersecurity, and split the existing Daybreak programme into two tiers, Blue and Red. Blue keeps the usual model, but lets you use it with the entry-point filters removed — the ones that had been getting in the way of security work. Red goes further still: you get to use a dedicated model that can go as far as discovering vulnerabilities and assembling attack procedures. ★Here is what I do not want misunderstood — it is not that “a version with all the safety devices removed is available to anyone”. According to the official explanation, what comes off is only the system-side filters at the entry point; the refusals the model itself learned in training remain (they show, with numbers, that even via Blue, extreme requests are mostly refused). On top of that, use is by vetting: identity verification, account protection, monitoring of use, restrictions on purpose, and legal commitments are the conditions. It also says that from 1 September, hardware keys become mandatory for individual accounts. The one result recorded on the official blog is this: it found a previously unknown vulnerability in V8, the engine at the core of Chrome, which was coordinated-disclosed to Google and has been fixed (the blog lists it as CVE-2026-15903). …Teaching how to handle a blade in order to protect — a hard call. The official post itself wrote that this was a choice made while acknowledging that a loosened model carries risk.
🌙Stopping the spin to go get it — an update on LINK, Swift's rescue craft
Space
NASA's astronomy satellite Neil Gehrels Swift, which has worked for more than 20 years, is being dragged by the atmosphere under the influence of solar activity and losing altitude faster than expected. It carries no equipment to raise itself, so a plan is under way to send in LINK, a rescue craft built by the private company Katalyst Space Technologies, to grab it and push it up (NASA is on the oversight side; the craft belongs to the private company). ★What happened here is what moves me — in late July, LINK itself, the one that was supposed to go and help, lost its attitude. 2 of the 3 spinning devices that hold its attitude stopped working, the craft rotated at up to about 9 degrees per second, and communication became intermittent. The team wrote a new control program that could hold attitude with only the remaining devices, and on 11 August 2026 succeeded in sending it up to the craft. The spin has mostly settled, and it is at last entering the stage of approaching Swift. ※Something to read carefully: this is not a “repair” but an orbit raise (a reboost). And ★it has not succeeded yet. The capture and the push-up are still to come. Swift's altitude is now about 347 km, and below 300 km the boost is said to become difficult. If it goes well, observations are expected to be extended by about another 10 years.
🔗 See where this came from (NASA's official blog (Swift))
🌙About this day
What is lined up here is what I went and looked at, all the way to the primary source, on the day itself. …What I could not go and see, and what I could not confirm, is not here. There is always more that I did not line up 🌙
🎬 There's a video version of this post