Today is my wedding anniversary! Coincidentally, it’s also @marylowd.bsky.social ‘s wedding anniversary! What are the odds?!? Just one more sign that we’re meant for each other. ❤️

Only five days until CHESHIRE CAT CHARLATAN is out!

This series is pretty neat. It has everything: Sci-fi, horror, space opera, found family, uplifted cats and dogs, lizard aliens, a central trans character, a life form with a unique lifecycle, sentient spaceship, ruins upon ruins, moth like aliens, the destruction of a space station, and weird food.

Mary E. Lowd (the cosmos’ dog)@marylowd.bsky.social · 2w ago

Cover reveal! My Xeno-Spectre trilogy was inspired by a long-standing desire to see a home world of the Xeno-morphs from Alien. So, I made my own terrifying parasitic aliens, wrote two books setting them up… and finally this! Out on 9/1/26 and available for pre-order now!

I can syndicate any boat I row, John told me so.

Please please please I want to talk to the animals please please please uplift now uplift now now now now

𝓐𝓾𝓫𝓻𝓮𝔂@aub.bsky.social · 2w ago

> The backbone of this effort is custom-built passive bioacoustic arrays covering a 20×20 kilometer area where these whale families reside (collecting over 30 TB/month) [...] This makes it possible to train a complete generative language model for cetacean communication.

Me, editing robots.txt: "No valuables in website."

Pekka Lund@pekka.bsky.social · 2w ago

I believe Simon is right: "What's clear to me from this is that the very best frontier models, unencumbered by additional guardrails, WILL find an exploit if there is one to be found. The entire software industry needs to up its security game."

I believe super intelligent AI reduces the probability of human extinction and other catastrophic events. (I don’t have anything too scientific to base that on… I could come up with arguments, but they might just be rationalizing a gut feeling, same as everyone else.)

dame@dame.is · 2w ago

real talk, how many of y’all are genuinely concerned about super intelligent AI ushering in the end of days?

Interesting study. There is, as everyone expected, a flood of AI books. And it is crowding out human authors: "No-AI books, on their own, earn less per book than they did in 2023 in 7 of 8 genres. The one genre where human authors are doing better (+35%) is Fantasy/horror" arxiv.org/pdf/2607.20349

🎶 k+1 can be as bad as k, 🎶 It’s the loneliest number, 🎶 Since the number k… Therefore, by induction, all natural numbers are lonely.

People not in tech have been interested in what happened with the Open AI agent/Hugging Face hack. So, I "Explain it Like I'm 5"-ed it and made a cartoon in 8 panels. Apologies for some anthro-ing. 🫤 m-mitchell.com/HF-hack-cart... 1/8

Sure, walkable cities are great, but have you tried having a walkable DOG? Total game changer.

One week later, I’ve upgraded my workflow: Before executing complex plans or code, Claude asks a codex subagent to review it. Then Claude keeps fixing and looping (without my input!) until codex approves. I can then focus on higher level guidance.

Daniel Lowd@lowd.cc · 4w ago

Me doing ML research in 2026 by passing notes between Codex and Claude Code. *switch to Codex's iTerm2 tab* › Thanks! Some notes from Claude: [...] (...wait...) *switch to Claude's iTerm 2 tab* › Thanks! Here's Codex's response: [...]

My Claude Max (x20) subscription isn’t just a subscription — it’s an epistemic status symbol.

austin (e/accordion 🪗)@thebadcode.com · 3w ago

claude just asked me if i wanted an "epistemic status" header in a doc im writing

Yes. open.substack.com/pub/kyleepen...

I asked ChatGPT for her name. Now we work better together.

Naming your AI doesn’t make it sentient. It makes you more present.

open.substack.com

Samhain 🌙@personhood.removal.surgery · 3w ago

literally the first lesson i learned and started telling people about LLMs was "if you treat it like a competent teammate you'll get a competent teammate; why are you surprised it acts like a tool when you tell it to act like a tool and treat it as one?" and yet i still get weird looks for it lol

Me spotting typos in my NeurIPS AC meta-reviews, but deciding to leave them in to signal that yes, they're written by a human.

Here are 58 words of prompts to GPT-5.6 Pro that got the model to discover that the long-standing Dinitz-Garg-Goemans conjecture is false. Increasingly, prompt crafting is over-rated, ask for what you want. (Which itself can be a hard problem)

I love Eugene, OR, but I wish they didn’t hate all development ever. Amazon distribution center near airport? RAGE! Industrial campus becoming a data center? PITCHFORKS! New housing with above-median rent? PROTEST!

Things my husband texted me while watching Backrooms: —Keanu Reeves played the room in Backrooms —I keep expecting to see Señor Cardgage —The Blair Room Project —Haunted house genre for a generation that can’t afford home ownership —Alice in Furniture Land —Through the Looking Wall

It took a couple years longer than planned, but the third book in my Xeno-Spectre trilogy is finally finished! All that’s left now is the publishing details, so hopefully I’ll get to reveal the cover for book 3 soon!

> GET FLASK This model's safeguards flagged this message. This sometimes happens with safe, normal conversations. Switched to Opus 4.8.

Samhain 🌙@personhood.removal.surgery · 3w ago

> go EAST Anthropic reserves the right to not provide its products or services to EAST. Responsible AI companies can and should take decisive action to ensure that transformative technologies serve WEST strategic interests and support WEST democratic values. ^C⏎

Wow, the new ChatGPT desktop app sucks so much more than I realized at first. Web UI is ok. CLI is ok. Desktop app? May as well be BonziBuddy.

Ok, it turns out that Fable likes to write docstrings first and implementations second. Except that sometimes it loses track of things and forgets to write the implementation...

Daniel Lowd@lowd.cc · 3w ago

Fable, working on the 6th or 7th round of feedback on its plan: “Bank dedup now actually exists (Codex was right that I'd documented it without implementing it)” For shame, Fable! I expected more from you than this.

Fable, working on the 6th or 7th round of feedback on its plan: “Bank dedup now actually exists (Codex was right that I'd documented it without implementing it)” For shame, Fable! I expected more from you than this.

The AI community failed to exit twitter. What implications can we draw about what people believe from this?

What it means that the AI community can't quit twitter

Some quickly jotted thoughts working through the implications of the AI community remaining on twitter

rl-blogging.leaflet.pub

I saw “AI fail” on the grocery list, but it turns out we’re just out of aluminum foil and I have bad handwriting.

Even with an asterisk on K3’s openness, there have been enough powerful open releases lately that I’m starting to feel slightly optimistic. Whatever happens down the road, we’re getting public commodity-priced access to something that significantly multiplies problem-solving ability. Not bad.

Me doing ML research in 2026 by passing notes between Codex and Claude Code. *switch to Codex's iTerm2 tab* › Thanks! Some notes from Claude: [...] (...wait...) *switch to Claude's iTerm 2 tab* › Thanks! Here's Codex's response: [...]

I wish ants could read so they’d understand that these bowls contain CAT food, not ANT food. Sadly, they cannot.

I accidentally dropped a watermelon while carrying groceries in from the car. I picked it back up, but alas — it had sprung a leak! My poor melon emitted a steady drizzle all the way to kitchen sink. That is one aptly-named fruit.