iiinova — where your digital self keeps evolving

dylan baker @dylnbkr.bsky.social · 28/09/2026

Looking behind the headline: 1. Guardrails = good. But when we let companies build their own guardrails, they are often avoiding the guardrails we NEED: ones imposed by third parties to serve the public interest. 🧵

13211

dylan baker @dylnbkr.bsky.social · 22/09/2026

lol I see I have been included in a list of west coast swing dancers to follow on bluesky. if you follow me from that, I hope you like uh... AI discourse 😅

030

dylan baker @dylnbkr.bsky.social · 22/09/2026

Reading the press about Jev and TypeSafe AI and it's wild to see ppl seemingly re-discovering what folks like @timnitgebru.blacksky.app have been saying this whole time. Yes, ML models should be well-scoped, structured, & testable! Being context-appropriate is, in fact, the most important thing!

A screenshot of text that reads: Large language models (LLMs) are designed to produce text for humans to read. When you need a model to make a judgment that your code will consume, that creates a mismatch: you are coercing a text-generation system into outputting structured decisions, then parsing the results back into something your code can depend on.

A screenshot that reads: "My experience is that Sutton’s bitter lesson is the tip of an iceberg of bitterer lessons: beyond compute and algorithms, there’s data and even picking the right task to do ML on." Below that is an image of a pink pyramid with four different vertical layers, which read top-to-bottom "algorithms", "compute", "data", and "doing the right task". Below that is the text: "The bitterest lesson in ML is that doing the right task > data > compute > algorithms."

25711

dylan baker @dylnbkr.bsky.social · 21/09/2026

cannot WAIT for this!!! If you want to really understand many of the issues with modern AI— the hyperscaling, the data theft, the labor exploitation—you need to understand ImageNet.

082

Reposted by dylan baker

The Distributed AI Research (DAIR) Institute @dairinstitute.bsky.social · 17/09/2026

DAIR's @dylnbkr.bsky.social weighed in on the mythical "kill switch" for @scientificamerican.bsky.social. Mitigating the harms of AI, as with any industry or technology, will instead require robust external regulation and safeguards. www.scientificamerican.com/article/what...

black text on a white background, reads: Dylan Baker, a research engineer at the Distributed AI Research Institute, goes further: the switch metaphor itself is the problem. “There’s such a complex ecosystem and web of interconnected technologies that we’re talking about that boiling this down to a single kill switch or ‘slowdown’ is reductive,” they say.

152

Reposted by dylan baker

Iris van Rooij 💭 @irisvanrooij.bsky.social · 15/09/2026

📣 Scholarship programme for applications by graduates in the Netherlands with a non-Western migration background and/or a recent asylum background. If interested to do PhD research in Computational Cognitive Science, feel free to reach out! Info about research lines in our lab in thread below. 🧵

nwo.nl

First round of Mosaic 3.0 now open for applications | NWO

The first round of the updated Mosaic 3.0 PhD scholarship programme is now open for applications. Mosaic encourages the recruitment of graduates from under-represented groups into Dutch academia by fu...

14533

dylan baker @dylnbkr.bsky.social · 15/09/2026

Roombas have SUPERHUMAN vacuuming abilities. & I don't want to alarm you, but when we taped a bunch of knives to one and gave it rocket boosters... it displayed emergent living room destroying properties. Please give me $100000000 to ensure rocket-powered knife roombas don't end human civilization 😔

1103

dylan baker @dylnbkr.bsky.social · 15/09/2026

Screaming this blog post from the rooftops fr Puts the last few weeks of press in very candid perspective: "There was some performative wrangling with the issues during the last week or two of press coverage, but generally terms like fascism, corruption, and deregulation were conspicuously absent".

000

dylan baker @dylnbkr.bsky.social · 11/09/2026

What is it that makes reading AI-generated text SO cognitively draining to read? It feels like a different kind of exhausting than reading poorly-written human text.

370

Reposted by dylan baker

Helen Czerski @helenczerski.bsky.social · 10/09/2026

As both an academic and a writer, I am repelled by the "this looks like a long document. Would you like me to summarize it for you?" that seems to appear at the top of every pdf viewer now. NO. The *entire point* of good writing (either books or journal papers) is that every sentence has a purpose.

1344044793

dylan baker @dylnbkr.bsky.social · 09/09/2026

"The work got done faster than I could have done on my own. And I barely had to think. While that sounds amazing if you're a person who profits from more work getting done, there are immense drawbacks to this" brettcodes.com/im-done-usin...

brettcodes.com

I'm done using AI

Why I'm stopping using AI for coding (and anything else) after using it earnestly for a year and coming to understand the harms it causes on personal, societal, and environmental levels.

040

Reposted by dylan baker

Ada Palmer @adapalmer.bsky.social · 03/09/2026

Britain is having its hottest summer on record, but one place has remained green. Brits in Holnicote have put rivers back onto their floodplains, restored wetlands, and released beavers that now build ponds and maintain the wetlands themselves. Countryfile buff.ly/CchGxpW #ShareGoodNewsToo

countryfile.com

Drought has ravaged the UK but this one nature site has been left untouched – this is how | Countryfile.com

Allowing nature to do its own thing could be the simplest solution to the UK's drought problem.

12498132

dylan baker @dylnbkr.bsky.social · 31/08/2026

"the 'rogue' frame [...] offer[s] up fantasies of a machine getting smarter. My worry is the intelligence that is retreating: the human intelligence that builds these systems, deploys them, and adopts them into workflows, then hides behind the results – then pins the blame on a system from nowhere."

0115

Reposted by dylan baker

Dan Hon @danhon.com · 31/08/2026

I think a reasonable position on the OpenAI huggingface thing is: * executing untrusted code is a bad idea * sandboxes frequently aren't, even with the best intentions * novel combinations of untrusted code are being generated at scale and being executed * consequences probably bad

38712

Reposted by dylan baker

Joseph Fasano @josephfasano.bsky.social · 24/08/2026

Back to school. ❤️

282566549

dylan baker @dylnbkr.bsky.social · 18/08/2026

Adding this to the list of reasons that synthetically generating probable text in situations where very specific word choice *really really matters*— e.g. law and medicine— is a terrible idea.

110

Reposted by dylan baker

Timnit Gebru @timnitgebru.blacksky.app · 13/08/2026

🔇 🔇 🔇 My book, Deep Unlearning, is now available for preorders! I don't know how people write so many books because this one took a lot out of me. It turns out, I had a lot to say 😂, with editors cutting out 50k of my words 🫣 ✴️ You can now preorder it at www.simonandschuster.com/books/Deep-U....

Cover of my new book, Deep Unlearning: The Rise of AI and The Radicalization of a Tech Idealist.

There are two concentric circles in the book, the left being circuits in black an white, and the right being a bunch of dots with orange accents. The cover background is cream. And the words are written in black except for "Un" which is in orange.

762186766

Reposted by dylan baker

Olivia Guest · Ολίβια Γκεστ @olivia.science · 08/08/2026

exactly 💯 on LLMs: they simulate the output we call language, not the cognitive capacity we call language!

492511700

Reposted by dylan baker

Alex Hanna @alexhanna.bsky.social · 07/08/2026

This statement from Codeberg maintainers, after their community voted to not use any code for LLM training and banned vibe-coded projects, is really good. blog.codeberg.org/protecting-o...

The development team of none

Using LLMs to work with your code gives you a kick of adrenaline. You can develop at a rapid pace, build things as if you had a large team. Only that you have none. In fact, you are (often) alone, working with a statistical machine that turns energy into code.

It seems like many ‘vibe coders’ don't realize that they don't actually have a community around them. They build projects as if they had, and spend resources accordingly. We see projects having a lot of code activity, heavy CI/CD testing, frequent and large release binaries. Sometimes, it feels like the amount of supported platforms exceeds the amount of actual users.

To us, it seems ridiculous to see projects with a single developer and virtually no users consuming as much or even more resources than some of the largest community projects on Codeberg, which operate frugal with CI/CD and storage resources. We do not believe it is reasonable for Codeberg to invest our precious donation money into hosting of large ghost projects.

5612170

Reposted by dylan baker

Dan "Mr. Bones and Me" Greene @dmgreene.bsky.social · 15/07/2026

It's union or bust, folks. This is expensive, we need taxes to make it happen. This is labor-led, we need need labor power to make it happen. Same dynamic in primary and secondary ed, in hospitals, in libraries, in childcare, etc etc

15512

Reposted by dylan baker

Anthony Moser @anthonymoser.com · 10/07/2026

it's ok to not be polite

Excerpt from I Am An AI Hater

To speak politely about AI, you put disclaimers before criticism: of course I’m not against it entirely; perhaps in a few years when; maybe for other purposes, but. You are supposed to debate how and when it should be used. You are supposed to take for granted that it must be useful somewhere, to someone, for something, eventually. People who are rich and smart and respected are saying so, and it would be arrogant to disagree with such people.

113540

dylan baker @dylnbkr.bsky.social · 10/07/2026

ooh, extremely open-source e-reader project!! "Open Book Touch is yours in a way most devices aren’t. The battery is one you can swap yourself. The library is one you load yourself. Nothing phones home, and nobody can reach in to delete a book you own." www.crowdsupply.com/oddly-specif...

crowdsupply.com

Open Book Touch

A pocketable, front-lit, open source e-reader — for every book, in every language

133

dylan baker @dylnbkr.bsky.social · 06/07/2026

yes!! it absolutely does not have to be this way!

02510

Reposted by dylan baker

Alex Hanna @alexhanna.bsky.social · 02/07/2026

A message for the class of 2026: We need more friction!

24460163

dylan baker @dylnbkr.bsky.social · 24/06/2026

"Ok I know about all these harms of generative AI. But what about when their coding tools are actually useful? I built this amazing thing that wouldn't have existed otherwise, what do you think about that?" I've had a lot of these conversations with socially-aware engineers and I have thoughts: 🧵

1142

dylan baker @dylnbkr.bsky.social · 21/06/2026

Reading a lot of speculative fiction where the preface is like "we must tell our own stories of hope, activism is world-building" and then the stories are like "yes, the plague after the fourth nuclear war had worsened the racism in Impoverishment City. But we never stopped dancing."

141

dylan baker @dylnbkr.bsky.social · 21/05/2026

Super excited to be speaking tomorrow at my alma mater! axl.hmc.edu/conference.h... The keynote I'll be giving is a talk I've been developing for quite a while— phrases like "AI Ethics" have come to mean a *lot* of different things. How do we actually orient ourselves in tech ethics conversations?

axl.hmc.edu

Conference 2026 | (Re)Imagining Education in the Age of GenAI

A two-day conference at Harvey Mudd College for faculty, administrators, and students. May 21-22, 2026 in Claremont, CA.

010

dylan baker @dylnbkr.bsky.social · 21/05/2026

Also big props to Yemariam Mamo (www.yemariam.com) and @paulinewee.bsky.social for this gorgeous design 🤩

yemariam.com

Yemariam Mamo

010

dylan baker @dylnbkr.bsky.social · 21/05/2026

So excited to finally announce this project! This has been such a labor of love from the whole team and I'm so pleased to see how it turned out 🥹❤️

13413

dylan baker @dylnbkr.bsky.social · 08/05/2026

"we know the mechanistic structure of such models because we designed and built them" thank youuuuu

0154

dylan baker @dylnbkr.bsky.social · 05/05/2026

This Wednesday! 🎉⬇️

000

Reposted by dylan baker

Remmelt 🛑 @artificialbodies.net · 30/04/2026

“the data center fight, with farmers and environmentalists and indigenous and working class people side by side, is an opportunity to grow and catalyze working class power. Fighting data centers is where good AI governance begins, not ends.” — @bcmerchant.bsky.social

bloodinthemachine.com

The data center rebellion is only the beginning

And it's precisely what democratic governance of AI looks like.

412232

dylan baker @dylnbkr.bsky.social · 29/04/2026

I wanted to share a blackout poem I made at the last Possible Futures workshop bc it captures how I feel a lot of the time: "I'm often accused of being an optimist, and it's true As angry as I may be As disappointed as I may be My heart is a strategic opening" [excavated from a Reagan speech]

A blackout poem on a torn-out page of a book. The book title at the top of the page is "Speaking My Mind", and the page number is 337. The visible text reads:

I'm often accused of being an optimist, and it's true
As angry as I may be
As disappointed as I may be
My heart
is
a strategic opening

2239

dylan baker @dylnbkr.bsky.social · 29/04/2026

New workshop next week, this time in New York! ⬇️

000

Reposted by dylan baker

jj skolnik @supremenothing.myatproto.social · 23/04/2026

public bathrooms that are accessible to all: a simple, necessary, revolutionary ask that we can all work on making happen right now

1043787

dylan baker @dylnbkr.bsky.social · 09/04/2026

I'm still thinking about one of my favorite tech demos from SXSW this year. Are you also stressed about the e-waste of single-use vapes getting tossed every day? Turns out, you can turn them into rechargeable, universal tool/e-bike batteries!! Other things are possible ✨ 2ndchancemnd.com

Text from Second Chance's website. It reads:

Our Solution
From Waste to Universal Power Packs
We transform discarded vape batteries into a product line of different universal power packs for consumer electronics that range from 12V to 48V Battery packs. These battery packs are compatible with major tool brands like Milwaukee, DeWalt, and Makita, along with various E-bike and E-Scooter makes and models. Our recycled battery packs deliver the same performance as OEM batteries while keeping harmful waste out of landfills.

01.
Collection
Partner shops collect used vapes in our specialized bins

02.
Processing
Batteries are safely extracted and tested for capacity

03.
Remanufacturing
Quality cells are assembled into tool battery packs

04.
New Life
Power tools, leaf blowers, and more run on recycled energy

A booth at sxsw labeled "Second Chance Material & Design". There is a table covered in power tools and large white 3d-printed batteries. There are also two clear plastic bins, one of which contains used vape cartridges (many look like Elf Bars), and another contains many circuit boards.

121

dylan baker @dylnbkr.bsky.social · 09/04/2026

Lots of us in the tech critical space get hit with the "do you even code bro" "do you even understand how this stuff works" all the time... meanwhile:

020

dylan baker @dylnbkr.bsky.social · 07/04/2026

I'll be running another virtual Possible Futures workshop as part of the Refugees in AI event ⬇️

011

Reposted by dylan baker

Timnit Gebru @timnitgebru.blacksky.app · 07/04/2026

🔇🔇🔇 On Friday April 17th (9am PDT/5pm CET/7pm EAT), we'll have a virtual event introducing our Refugees, Migrants and AI research program which centers the lived experiences of refugees and migrants and puts power back into their hands🧵 ☘️☘️☘️ Register at www.eventbrite.com/e/refugees-a...

eventbrite.com

Refugees and AI

Centering the lived experiences of refugees and their power building work in fighting against surveillance technologies.

210067

Reposted by dylan baker

Ana @anthrotechana.bsky.social · 07/04/2026

PNW folks, please consider joining us at PDX 350 for an in-person art-making workshop. @dylnbkr.bsky.social and I will lead this event, designed to prompt us to imagine alternative technofutures and be in community at such difficult times. Registration/Info: www.eventbrite.com/e/dair-data-...

eventbrite.com

DAIR + Data & Society Present: Imagining Possible Futures, Portland

Imagine alternative technofutures with the Distributed AI Research Institute and Data & Society in this free, in-person art-making workshop.

142

Reposted by dylan baker

Chris Plante @plante.bsky.social · 07/04/2026

If you need a 1-hour break from doomscrolling, I recommend this interview with Danielle Brathwaite-Shirley, the creator of BlackTransArchive.com. Art can challenge us, hold us accountable, AND still give us hope. postgame.substack.com/p/video-game...

04811

dylan baker @dylnbkr.bsky.social · 07/04/2026

Got another workshop coming up this weekend in Portland! ✨

022

dylan baker @dylnbkr.bsky.social · 31/03/2026

File under "a lot of damage isn't irreversible and other things are possible" ⬇️

064

Reposted by dylan baker

Prof Rachel Oliver @profrachelgan.bsky.social · 23/03/2026

I had an experience in a meeting today which showed how old school note taking can actually get you ahead of the game. The meeting involved a document that's about 400 pages long. It's exactly the kind of thing that AI fans would suggest should be summarized by AI. Instead...

16617

Reposted by dylan baker

Elf M. Sternberg @elfsternberg.bsky.social · 23/03/2026

Game developers share their deliberately crude placeholders to make the point that AI illustration for placeholders isn't offensive, it's poor practice because it could make it into production. kotaku.com/crimson-dese...

kotaku.com

Game Devs Reveal All Their Ugly Placeholder Assets Made Without AI

Crimson Desert's AI placeholder art controversy has led to many devs sharing sillier temp assets made by humans and MS Paint

3438

Reposted by dylan baker

Ana @anthrotechana.bsky.social · 19/03/2026

So happy to bring this event to Oregon with @dylnbkr.bsky.social 🥹

021

Reposted by dylan baker

DAIR Possible Futures @dairfutures.bsky.social · 19/03/2026

New in-person workshop incoming, this time co-hosted with @datasociety.bsky.social in Portland, OR! Saturday, April 11, 1pm-3pm at 350 PDX. For tickets and more info, head over to Eventbrite ⬇️ www.eventbrite.com/e/dair-data-...

A promotional image for the workshop.
Imagining Possible Futures
Portland
Saturday, April 11, 2026
1pm-3pm
3639 N Mississippi Ave
Portland, OR

0126