iiinova — where your digital self keeps evolving

Congrats! Hope it's a good busy. I definitely wonder how some folks On Here manage to post so frequently, given that they're also fully employed. My days have never been so relaxed as to enable fully online behavior.

111

For a platform that believes in controlling your own experience and community moderation, @bsky.app could use a lot more speedbumps for preventing poor behavior

Enforcement of anti-brigading rules after the fact is a good thing, but features to control your experience are even better. (I feel like that's part of the notional platform philosophy anyway?)

Portalrealm @portalrealm.bsky.social · 28/09/2026

I continue to think bsky should have speedbumps built into the design, e.g. - A modal warning if QTing a much smaller account - A modal warning if your reply seems extra-spicy - Configurable triggers that auto-close replies/quotes if you get too many interactions too fast etc.

260

Reposted by ajdecon

Today, 39 years ago, a beautiful ship glided through space.

17116

oh god I’m going to have to steal that image, it’s amazing

Or to phrase it differently: Jensen cares more about NVIDIA's success than anything else in the world. If you work at NVIDIA, he expects the same from you. Whether or not that's a good trade is up to you! It's in many ways a very fun time, it's just also rather prone to burnout.

To be clear, I spent the majority of my career (so far) at NVIDIA. The company really is very good to its employees in terms of pay, benefits, time off, and even being human at work. The trade is that NVIDIA also demands a much higher level of focus while *at* work than most other workplaces.

* Re: influencing politics, NVIDIA does of course lobby the government, but seems to do so in very boring "business-friendly environment" ways

100

More seriously -- NVIDIA wants to be a company where you can "do your life's work", and I think that philosophy starts at the top. AFAICT, Jensen doesn't spend time fantasizing about AGI or trying to influence politics* or even buying toys. He just wants NVIDIA to be as successful as possible.

I mean of course Jensen cares about more than selling GPUs. NVIDIA also makes CPUs and network gear after all.

bryan newbold @bnewbold.net · 28/09/2026

Reposted by ajdecon

Hello World from the PLC organization (plcred.org)! The org was soft-announced back in March at AtmosphereConf. We've been working through logistics step by step and now have a public web presence we can point to.

blog.plcred.org

First steps of the PLC organization

1425165

Related: at the point where your colleagues chat about not wearing swag in public, it’s probably a good idea to consider making a change.

030

🤷‍♂️ don’t know how an interviewer can stop you (unless in person, and the smell gives you away!) but I don’t think it’s an edge case important enough to police folks’ hydration in interviews

100

Because this discourse is still going, I feel like I should say: If I’m interviewing you, feel free to have an iced coffee. Or a regular coffee. Or a bottle of water. Or whatever (non-intoxicating) beverage or snack helps you perform. Treat people like adults who can manage their needs, please.

Pay attention to Cat’s work, she’s doing great stuff in this area

Those histories are also honestly helpful to practitioners, given how often engineering is historically contingent. So frequently, “why does X work that way?” traces back to “because this team 30 years ago made a decision based on where they trained or what was cheapest at the time”

021

Entirely separate from the quotes on general liberal arts, I’m really sad about how little science/eng degrees teach the history of the fields themselves. The histories of fields like physics, computing, structural engineering, genetics, and so on are all fascinating!

1102

That seems like it should be an additional course rather than replacing the existing topics, to be clear. Everyday engineering ethics is still probably the most relevant bit for most working professionals. Which leads to the increasing problem of “so many required courses”… etc

020

AFAICT, most undergrad engineering ethics education focuses on “nuts and bolts” topics. E.g., avoid conflicts of interest, don’t compromise safety to reduce cost, no bribes, that kind of thing. Those are all important! But it does feel like more coverage of societal impact is needed these days.

Jess Miers 🦝 @jmiers230.bsky.social · 27/09/2026

I’m similarly angry that company names with LOTR references are now red flags

030

Reposted by ajdecon

Google brought back their retro logo for their 28th today. They should bring back their retro principles too.

29214

ajdecon @ajdecon.org · 26/09/2026

There is very little as frustrating in (online) chess, as mis-clicking a move badly, and then fightin a losing battle for the rest of the game. Literally had the phone slip in my hand and moved my queen into the worst possible position. 😮‍💨

ajdecon @ajdecon.org · 26/09/2026

This is a good description of something that frustrates me regularly. So many folks seem to ignore the actual goals of whatever effort they’re part of, in favor of applause from people they like.

044

ajdecon @ajdecon.org · 26/09/2026

I’m seeing iced coffee discourse here and how did this migrate from threads

040

ajdecon @ajdecon.org · 25/09/2026

I mean totally fair. TBH part of it also comes from the experience of using them. Better results if your input resembles talking to a person than machine instructions.

ajdecon @ajdecon.org · 25/09/2026

In certain cases, sure. I talk about code generation or text generation a lot. Tool use makes it awkward. “The LLM generated code which resulted in many web requests to search engines with keywords about trains”? Nah, the LLM went and researched trains.

ajdecon @ajdecon.org · 25/09/2026

This is basically where I am. If I say the LLM thought about something, that doesn’t mean I think it’s doing the same thing as a human. It’s just a lot less tedious than saying “the LLM appeared to process text about X in a way that produced a particular result”

141

ajdecon @ajdecon.org · 24/09/2026

*sigh* This is a very Jensen thing to say. He spends 16+ hours a day running NVIDIA, because to him it's the most fun thing he could be doing. This shows up in other interviews too. Of course he doesn't remember his zip code, he's restructured his life so he doesn't have to.

ajdecon @ajdecon.org · 22/09/2026

Ha! Congrats @segyges.bsky.social on your bot making the @bsky38.com list

“User was blocked for the post” at #22 on the list

ajdecon @ajdecon.org · 20/09/2026

I’ve seen this happen in two-author situations at least. (Paper A has author A as first author, Paper B for B) Often it’s e.g. an experiment-focused paper and an analysis-focused paper, submitted to different journals. I haven’t personally seen it for larger numbers but feels possible 🤷‍♂️

030

ajdecon @ajdecon.org · 20/09/2026

I don’t disagree, but also it’s just exhausting to see so many people unwilling to acknowledge uncertainty.

000

ajdecon @ajdecon.org · 19/09/2026

Oxide could probably do decent business with a public swag store

020

ajdecon @ajdecon.org · 19/09/2026

What this says to me is that most datacenter regulations are more responsive to constituent sentiment (which is fair!) than to any specific DC build of concern. - 300+ datacenter bans or moratoriums in the US - Only 3 projects actually meaningfully impacted so far

newsletter.semianalysis.com

Everyone Says Datacenter Moratoriums Are Killing the US Buildout. We Mapped All 300 of Them

20GW sits inside a restricted local boundary, 1,525MW actually slips, 2.3GW nationwide including New York

022

ajdecon @ajdecon.org · 18/09/2026

It brings to mind the obsession with “gray goo” scenarios when nanotech was a big thing in science. Not so much a realistic scenario as “this would make an AMAZING sci-fi movie”.

031

ajdecon @ajdecon.org · 17/09/2026

this is the correct response to a suspension

040

ajdecon @ajdecon.org · 16/09/2026

I really hate that I understood all of this immediately without deep thought or googling

020

ajdecon @ajdecon.org · 16/09/2026

IIRC, the Three Laws were basically a hook that Asimov used to write often-playful short stories about logic problems, and how to find loopholes in rules. Fun, but hardly the basis for serious safeguards on AI.

Maria Antoniak @mariaa.bsky.social · 14/09/2026

Reposted by ajdecon

We're hiring in Computer Science at the University of Colorado Boulder! ☀️⛰️ Machine learning and NLP people, please apply!

05531

As usual, Charity's thoughts here are worth reading

Disclosures: I’ve worked professionally on… - GPU / machine learning infrastructure - Storage systems for social media - Nuclear weapons simulation infrastructure - And now, quantum computing infrastructure All of that will inform my views on, like, everything. 🤷‍♂️

000

I would sincerely love to find actionable advice on this beyond “tell your parents to ask you about each interaction”. Even at the level of “ignore it all until you get a letter in the mail”, which *feels* right but may not be. My parents are fairly savvy, but I worry.

1120