iiinova — where your digital self keeps evolving

Tolki @tolki.dev · 28/09/2026

Tolki @tolki.dev · 20h

Haasta la vista

011

Man the Japanese fascists having managed to convince themselves Japanese news outlet have a left wing bias has to be the craziest shit ever How can you be so far removed from reality and actual facts?

140

Tolki @tolki.dev · 28/09/2026

Are we getting a clean 4k release soon? It was hell to find in good quality a few years ago

000

Reposted by Tolki

Nick Kapur @nickkapur.bsky.social · 24/09/2026

Japanese used bookstores are reporting a massive uptick in bulk purchases of used Japanese books by overseas buyers, seemingly without any regard to their contents. The only logical conclusion is that these books are being stripped, scanned, and discarded to train AI models.

416896

Reposted by Tolki

Agence France-Presse @afp.com · 21/09/2026

🇺🇸 Le gouvernement américain a expulsé plus de 25.000 personnes dans des pays tiers, aux termes d'accords secrets passés avec 35 pays pour qu'ils accueillent des personnes n'ayant aucun lien avec eux, révèle une enquête journalistique mondiale. ➡️ u.afp.com/Sjjy

62624

Tolki @tolki.dev · 20/09/2026

I think today’s pretty good compared to what we’re getting tomorrow… Still could enjoy a stroll in the park at least!

011

Reposted by Tolki

Alasdair Beckett-King @misterabk.bsky.social · 19/09/2026

Well, of course those hippy Marxists at the Financial Times would say that.

026159

I think there’s enough of a power differential for reverse grid Monaco sprint to be a shitshow for the ages I just want to see Bottas start from pole

Sprint in Monaco is pure evil ah ah ah OR we do it with reverse grid and have some fun

120

But liability on the model output itself seems a bit too broad and pretty close to a full ban on LLMs, which I don't think is a realistic goal or approach at this point

000

While I understand and agree with the general idea, I feel like it's more important to make LLM companies liable on the training side: full disclosure of content used in training + modifications made to it They should also be made liable for the tools they operate around LLMs like chatbots

200

> The same way we were accountable for the software we write The MIT licence used by Godot literally contains: IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM (...) ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE

220

Tolki @tolki.dev · 14/09/2026

How would that work for open weights models?

100

Well you can go up to four, the katakana and the hiragana also are pretty different!

000

Tolki @tolki.dev · 12/09/2026

And the government will continue doing jack shit about the root causes (cost of living crisis due to wealth inequality and climate change)

120

Tolki @tolki.dev · 11/09/2026

He’s talking about you?

Tolki @tolki.dev · 09/09/2026

Achetable au Japon ? 👀

Tolki @tolki.dev · 09/09/2026

This ain’t F-Zero

000

Reposted by Tolki

Sandrine Rousseau @sandrousseau.bsky.social · 09/09/2026

Les choix des photos par les rédactions c’est pire que chez Disney ou Marvel. En une image tu sais qui est gentil et qui est méchant.

Photo utilisée par le parisien avec 9 visages roussel/philippe/glucksman/attal/melenchon/Bardella/ruffin/lepen/tondelier
Tous sont souriants … sauf 1 Jlm qui grimace

Pareil Jadot rousseau
Jadot sourit et moi je fais la gueule

49985251

Tolki @tolki.dev · 09/09/2026

Well it’s the dems that wants their votes, right? So maybe they could do like… anything, instead of their whole platform being just “hey at least we’re not Trump”

131

Tolki @tolki.dev · 08/09/2026

What’s wrong with the penguin

030

Reposted by Tolki

Fabrice Arfi @fabricearfi.bsky.social · 07/09/2026

Matignon demande à la justice de traquer les sources de journalistes en invoquant « la vie des affaires » et « les intérêts supérieurs de l'État » alors qu'il est question d'informations publiées sur l'élaboration... du budget de la nation. Insensé.

461106639

Tolki @tolki.dev · 08/09/2026

Btw Jubeat is too mainstream pls retcon it to Wacca

Tolki @tolki.dev · 08/09/2026

Hey I have that triforce t shirt Nintendo distributed it at Japan Expo 2009

111

That looks hilarious, hope they scale it up!

Most theories of consciousness until a few years ago would have classified LLMs as conscious (Functionalism, global workspace, higher order, attention schema, ...) I fully agree they are *not*, which is why they make us question our definitions

320

Yes, and what’s human intelligence if not that? We operate based on imitation and mimicry. New ideas don’t come from thin air but usually from reframing existing information. LLMs are trained similarly, in that they’re given massive volumes of input and associated actions and try to mimic them.

100

AGI is indeed an empty buzzword but I still feel like you’re a bit too dismissive of LLMs and particularly the philosophical implications Like, actually philosophy like defining conscience or intelligence

550

I’ll take two

000

Maybe some weird ass system where 2 teams play PvP in a round while another plays PvE that impacts what happens on the PvP map, rotating every round

Man you just made me want a SC CS clone 3 races, 3 teams per map lol

Tolki @tolki.dev · 02/09/2026

I think you need to have friends to understand the joke

000

Have you tried not being poor smh

Tolki @tolki.dev · 02/09/2026

That’s what you get for not buying my keyboards…

Jay Van Bavel, PhD @jayvanbavel.bsky.social · 01/09/2026

Reposted by Tolki

I just finished reading "Palaces for the People" which explains how locations, like the library, are not only tools for learning but also for connecting people and bridging divides. The future of democratic societies rests not simply on shared values but on shared spaces. I highly recommend it!

2226

Tolki @tolki.dev · 28/08/2026

It’s 45 minutes from my place, might make a jump on Monday 👀

000

Reposted by Tolki

Jay Van Bavel, PhD @jayvanbavel.bsky.social · 28/08/2026

Is doomscrolling bad for you? A meta-analysis of 42 studies (N=46,912) finds that consuming short-form videos is linked to increases in inattention & impulsivity; reductions in working-memory span & self-regulation; elevations in anxiety, depression & stress; and a surge in addiction symptoms.

16456174

Tolki @tolki.dev · 27/08/2026

Not sure why they picked the League of Legends midlane logo but why not