THE SHIPLOG_

MayaThis is The Shiplog, Friday, October 9. Okay, picture this: seven Adobe-style apps, free, open source, all rewritten from scratch in Rust, and the claim is it was basically one developer.

TheoAnd that's not even the only story today where one person plus AI looks like a whole team. OpenAI also just dropped hundreds of math manuscripts at once.

MayaWe've also got a Starlink license fight in India, a South Korean bank hack tied to a single person, and what AI spending is doing to financial statements. Let's get into it.

TheoSo, ArtCraft. The site getartcraft.com lists seven free, open-source creative tools, all built in Rust. Image editing, vector illustration, video, photography, PDFs, motion graphics and page layout.

MayaWhich is basically the Adobe lineup, covered end to end. They're presented as reimplementations of Adobe products. Wait, how does one developer even attempt that?

TheoThe claim is they used Claude Opus 5.5 to re-create the apps, including Photoshop, Premiere and Lightroom equivalents, and then ported everything to Rust. And they reportedly believe they can match Adobe's features within months.

MayaWithin months is a big promise. But I see why people care. Adobe's all-apps plan runs $840 a year, so a free alternative lands differently. Still, matching a feature list is not the same as earning a professional's trust.

TheoRight. And Eric S. Raymond points to a related project, photocraft, a clean-room open-source Photoshop reimplementation. His guess is the recipe was decompile the app, convert it into a spec, then prompt an LLM to write Rust.

MayaHuh. So if that works, the code stops being the moat. He argues that threatens closed-source software in general. But the obvious tension is, can you call it clean-room if the spec came from decompiling the original?

TheoExactly the question. The notes don't settle it, but if you sell software, you should be watching this.

MayaOkay, shifting to India. Actor and politician Kamal Haasan is responding to Elon Musk on satellite internet, and he opens by agreeing with a piece of it: no billionaire should dictate a nation's policy, Indian or American.

TheoRight, he's not picking a side between billionaires. His position is that India should welcome competition from Indian and global providers, whoever offers the most practical and affordable last-mile solution for people in remote regions.

MayaAnd affordability is the center of it. He says in countries with lower per-capita incomes, satellite internet has to be affordable too, because high subscription fees and equipment costs could put it beyond ordinary households.

TheoThen he cites a CAG audit finding that BharatNet is struggling to meet its objectives. So the subtext is, the existing public effort isn't closing the gap, so the door shouldn't be shut on alternatives.

MayaI like that framing. It's less about Starlink and more about whether a household in a remote area actually gets connected, and what they can afford once it arrives.

TheoOkay, big one. OpenAI is releasing a broad range of new mathematical results produced by an internal frontier model. And it says it consulted the Institute for Advanced Study's Advisory Group on Mathematics and Artificial Intelligence on how to release them.

MayaThat consultation detail matters. They're signaling they know dropping this on the math world is delicate. Materials are on GitHub. So how big is big?

TheoReportedly 722 manuscripts, grouped into 372 families of results, drawn from about 4,000 research problems. And the average was about three hours of ChatGPT Pro compute per result.

MayaThree hours per result. That reframes the cost of a theorem. What about the actual claims?

TheoHighlighted ones include a zero-free half-plane for the zeta function and a quasi-Riemann hypothesis advance. But those remain to be independently assessed, so treat them as claims, not settled math.

MayaAnd the human side is already loud. NYU math professor Tristan Buckmaster says the release wiped out entire research programs and damaged early-career mathematicians' careers.

TheoHere's the tension. When results arrive in bulk, checking them becomes the bottleneck, and the people most exposed are the ones whose reputations depend on a few key results.

MayaThis episode is brought to you by Riti. It's AI virtual try-on and a styled fitting room for Shopify apparel stores. Riti styles shoppers' pieces into complete looks, rendered on their own photo.

TheoThe whole look goes into the cart in one tap, with sizes picked per piece. It installs as a theme app block, no code. Every plan starts with a 30-day trial and 50 try-ons. Find it at rititech.app.

MayaOkay, back to the news.

TheoSecurity now. Last week some of South Korea's biggest banks were hit by a cyberattack, and a CrowdStrike report suggests the entire thing may have been done by a single person.

MayaWith a stack of tools: an open-source AI penetration tool called ARTEX, plus DeepSeek v4.1-Flash, GLM-5.3, Grok 4.6 and Claude Code. Open source and commercial models, mixed together.

TheoAnd here's the part that got me. Researchers reportedly identified him because he left directory listing enabled on his server, which exposed his Claude Code session files.

MayaSo the AI multiplied his reach, but not his operational discipline. That's almost comforting. Still, the claims are as relayed, and the bigger point is that one person could plausibly hit major banks.

TheoBack to Starlink, because the other side of that India debate is Musk's own statement. He says Starlink is licensed in over 165 countries and has spent five years complying with every single law and requirement of India's government.

MayaSo his question is, why still no license? And then he asks, is Ambani the real boss of India? He was responding to an allegation from Dr Aniruddha Malpani that the government protects crony capitalists over citizens.

TheoWorth being careful there. Those are accusations and rhetoric, not findings. But the plain question, five years of compliance and no license, is a fair one to ask.

MayaAnd it ties back to Haasan. Both agree policy shouldn't be dictated by a billionaire, but they land differently: one says open the market to competition, the other asks why his company is still waiting.

TheoLast one, and it's a good closer. Aswath Damodaran has a new blog analysis of how trillions of dollars in AI capital spending have flowed through financial statements.

MayaHis point is that it's not just markets being roiled. The spending is distorting book capital, earnings and cash flows, so the metrics people use to value companies may be skewed.

TheoAnd he frames the cause as hope of riches. That's the tension: spending driven by expectations, while the accounting numbers we rely on to judge it get bent in the process.

MayaSo if you're reading valuation ratios on AI-heavy companies, you may be reading a distorted picture. That's worth keeping in mind next time earnings come out.

MayaThat's The Shiplog for Friday, October 9. You can read every story at shiplog.news, and please subscribe.

TheoThanks for listening, everyone. Have a great weekend.

MayaSee you next time.

Generated by AI from the paper's summaries; voices by ElevenLabs. Check the original posts for details.