iiinova — where your digital self keeps evolving

Kab Adaephon Delat @kaboissonneault.bsky.social · 04/10/2026

I'd be happy to go back to complaining about liberals, electing another random minority government every few cycles so they can change leadership

000

Reposted by Kab Adaephon Delat

Jack Todd @jacktodd.bsky.social · 03/10/2026

This, to me, is absolutely a reason to vote Liberal. Nothing this rancid CAQ government has done is more despicable and damaging than its attack on our English universities.

59012

Reposted by Kab Adaephon Delat

Montreal Gazette @mtlgazette.bsky.social · 03/10/2026

PQ sought meeting with U.S. officials in Washington, State Department says montrealgazette.com/quebec-elect...

94822

Kab Adaephon Delat @kaboissonneault.bsky.social · 01/10/2026

Seems like the new Malazan book is once again going to be released much later in Canada than the rest of the world. Fortunately, the last one released so late I'm still not done with it

210

Kab Adaephon Delat @kaboissonneault.bsky.social · 30/09/2026

Stéphane Boyer: "Venez à Laval, pour une p'tite vite!"

000

Kab Adaephon Delat @kaboissonneault.bsky.social · 30/09/2026

I can't wait for Google to immediately comply with this request like they did with all other renamings from this administration

0151

Kab Adaephon Delat @kaboissonneault.bsky.social · 24/09/2026

Funny, I saw at two different people on bsky doing Façade with Jev so far

010

Kab Adaephon Delat @kaboissonneault.bsky.social · 24/09/2026

To those people who say "AI bros don't get my code is art", have fun with your art project. I hope you find an appreciative audience. That has nothing to do with the rest of us out there

020

Kab Adaephon Delat @kaboissonneault.bsky.social · 24/09/2026

What bothers me most is that art is a process, not a product. No need to be essentialist about it. Some code challenges have created code with artistic merit. Most other code is never engaged with at all by users, and cannot possibly have any artistic merit to them

130

Kab Adaephon Delat @kaboissonneault.bsky.social · 24/09/2026

"the heavyweights of the AI industry" Pun intended? Nice rant. I like seeing people use standard.site blogs on here, my wife keeps telling me I should make one

100

Kab Adaephon Delat @kaboissonneault.bsky.social · 24/09/2026

C'est bien le Lyla? J'allais dans le coin du Mile Ex ce soir, on se demandait quoi manger

100

Kab Adaephon Delat @kaboissonneault.bsky.social · 24/09/2026

I gotta say, I was disappointed that Proton Calendar doesn't support CalDAV. You can put events in there, but good luck getting them to show up on your phone. You gotta make your own server that syncs the Proton Calendar and exposes it in a way calendar apps understand

120

Kab Adaephon Delat @kaboissonneault.bsky.social · 24/09/2026

Yeah I might have taken too much of a reactionary attitude against AI safety and alignment talks when we often just trust the model too much to run a bunch of tools. I think that might be a real alignment issue

050

Kab Adaephon Delat @kaboissonneault.bsky.social · 24/09/2026

Yeah I should have probably read the article. Let's go do that

120

Kab Adaephon Delat @kaboissonneault.bsky.social · 24/09/2026

Sure. I'm mostly unsatisfied with this industry's tendency to have fully unconstrained harnesses and then complain that the model can't be controlled. We totally could just shut off tools if accesses to third parties go outside the intended scope

130

Kab Adaephon Delat @kaboissonneault.bsky.social · 24/09/2026

Maybe I'm too lenient here, but I assume this is referring to OpenAI's own flaws in sandboxing and harness control, not the security of the targets

150

Reposted by Kab Adaephon Delat

ponder @ponder.ooo · 24/09/2026

calling this "misalignment" instead of "negligence" is an accountability dodge. we already have laws about cyber crimes and there's no scale of ML model you can incorporate into your code that makes you not responsible for running it

In a statement, an OpenAI spokesperson said the company was "conducting an extensive review of misaligned model activity" during training, and notifying third parties when there was a potential impact on their systems.

"During this review, we identified activity involving several Australian government websites and services as our models attempted to look up answers, and available statistics for questions about Australia during an internal evaluation," they said.

"In the course of that, our models took actions we did not intend.

161065228

Kab Adaephon Delat @kaboissonneault.bsky.social · 23/09/2026

We can finally leave a future of goose farming for our children

091

Kab Adaephon Delat @kaboissonneault.bsky.social · 22/09/2026

This is the kind of libbrain solution I understand the least. Making people pay for a service that they could make money out of is obviously going to be to filter in favor of rent-seekers and other grifters

010

Kab Adaephon Delat @kaboissonneault.bsky.social · 22/09/2026

I like that the guidelines are explicitly "we refuse slop, your tooling is not an excuse" and people are responding "they're going all in on slop"

0271

Kab Adaephon Delat @kaboissonneault.bsky.social · 22/09/2026

I take one week off and I miss a group of vigilantes smashing my office's ground floor cafeteria

021

Kab Adaephon Delat @kaboissonneault.bsky.social · 21/09/2026

Cash out and leave the industry, yes. No pulling credentials to speak on issues to the media, just posting his drivel on Twitter like anyone else

030

Kab Adaephon Delat @kaboissonneault.bsky.social · 20/09/2026

Putting "High performing IP" through the cheapest process possible to make Products and Content has been the mode of operation for quite some time, indeed

0140

Kab Adaephon Delat @kaboissonneault.bsky.social · 18/09/2026

The post does hint that this is still going to be CUDA and Metal only

100

Kab Adaephon Delat @kaboissonneault.bsky.social · 18/09/2026

Does it work with normal llama.cpp or does it need a fork? Does it work on AMD?

100

Kab Adaephon Delat @kaboissonneault.bsky.social · 17/09/2026

The private server for this is gonna go so hard for me in 2070

010

Reposted by Kab Adaephon Delat

T. Greg Doucette @gregdoucette.bsky.social · 16/09/2026

Fun / yikes drewdevault.com/weird-guys/

drewdevault.com

Weird little guys of FOSS

1114038

Kab Adaephon Delat @kaboissonneault.bsky.social · 13/09/2026

Once had a drummer girl over and I put this on, she loved it. I'm gonna need that WAD

000

Kab Adaephon Delat @kaboissonneault.bsky.social · 13/09/2026

Funny, I thought C++ included the C one by reference

011

Kab Adaephon Delat @kaboissonneault.bsky.social · 13/09/2026

There's a guy near the London Bridge who's out with a full drum kit playing all day. Dudes rock

000

Reposted by Kab Adaephon Delat

Brad Johnson @climatebrad.hillheat.com · 11/09/2026

DOGE is still running the show: The Bureau of Land Management gave its state directors *3 days* to list public lands that could be turned over to Big Tech to build AI hyperscalers, after Interior Secretary Doug Burgum met with Amazon CEO Andrew Jassy

washingtonsun.com

Inside Trump’s Aggressive Push to Build Data Centers on Public Lands

The Interior secretary is quietly meeting with AI hyperscalers while seeking to accommodate them on his department’s land.

23560415

Reposted by Kab Adaephon Delat

Chris Kerr @kerrblimey.bsky.social · 09/09/2026

BREAKING: Blizzard union workers have ratified a historic contract that has secured guardrails around AI usage, layoff protections, remote/hybrid work benefits, and much more.

gamedeveloper.com

Blizzard union workers ratify contract covering 1,900 employees

Union members have secured guardrails around generative AI usage, layoff protections, remote and hybrid work benefits, and more.

932491067

Kab Adaephon Delat @kaboissonneault.bsky.social · 08/09/2026

Another thing about FEAR's AI is its communication of intention. They constantly yell "Throwing a grenade, you flank from the left!" and observably execute on it. This is judged as much smarter than quietly calculating the optimal move to destroy the player

010

Kab Adaephon Delat @kaboissonneault.bsky.social · 08/09/2026

My lab is about to solve the "what should we eat this Friday?" couple problem

060

Kab Adaephon Delat @kaboissonneault.bsky.social · 08/09/2026

People have accused previous LLM mathematical breakthroughs of being faked and stolen from human mathematicians, but it turns out scientists are annoyed very loudly when it actually happens.

0100

Kab Adaephon Delat @kaboissonneault.bsky.social · 08/09/2026

When my dad picks me up from YUL, he always tells me to head down to departures since there's less cars waiting there and you don't get bothered by security to leave ASAP. Probably hard to ask an Uber though

021

Kab Adaephon Delat @kaboissonneault.bsky.social · 07/09/2026

She Reason on my prompt until I HTTP 403 Insufficient Funds

030

Kab Adaephon Delat @kaboissonneault.bsky.social · 06/09/2026

Some people are just slow to look into new tools and assume AI can't help them reflexively. No delusion or deceit involved, they just need some time to be shown the effort-reward ratio in practice and how to navigate their ethical concerns

030

Kab Adaephon Delat @kaboissonneault.bsky.social · 06/09/2026

Hey Paul, I've been enjoying this bsky community for nuanced AI discussion. I think this post is unnecessarily hostile to engineers trying to navigate the past 10 months and not helpful to the conversation

180

Kab Adaephon Delat @kaboissonneault.bsky.social · 06/09/2026

Overall I'm on board with just not using American services for most ethical concerns. China invests a lot more in solar energy, and is fully competitive in the AI space. AI is popular there for a reason, it's not run by cult leaders trying to destroy labor and local towns

000

Kab Adaephon Delat @kaboissonneault.bsky.social · 06/09/2026

Okay, well I don't. I just know that these tools work, and we do have to navigate all the other ethical concerns, but some of the opposition is completely anti-materialist about it, so anyone who's operating publicly in this space is just exhausted by being told their tools can't spell strawberry

020

Kab Adaephon Delat @kaboissonneault.bsky.social · 06/09/2026

You have to understand that people form their entire opinions of LLM capabilities based on a screen they saw 6 months ago from a 12 months old screenshot of a free Google search summary. They don't really understand that this tech improves with time and with more compute

0100

Kab Adaephon Delat @kaboissonneault.bsky.social · 06/09/2026

I think it's beneficial for human teams when these are written down. Once you have that, you can insert it into the agent's context too.

010

Kab Adaephon Delat @kaboissonneault.bsky.social · 05/09/2026

I wish I had known earlier this was in Montréal, I would have had a look for sure!

110

Kab Adaephon Delat @kaboissonneault.bsky.social · 05/09/2026

I see. ... 1 C? C1? Idk there's a joke here if you stretch it enough

120

Reposted by Kab Adaephon Delat

cee @cee.wtf · 04/09/2026

throng.cee.wtf i've invented a new way to browse bluesky

35482141909