Okko Ojala @[email protected]

Developer and sw architect.

[Okko Ojala @[email protected] @okko]

Blogissa: THL:llä ei yhtäkään asiakirjaa koronaviruksesta kouluissa lukuvuonna 2021–2022, seuranta lopetettu ilman päätöstä. Deltan ajalta ei siis yhden yhtä asiakirjaa koulualtistumisista tai jatkotartunnoista. https://t.co/hHZvLP0rR8

[Okko Ojala @[email protected] @okko]

@LiskoTouko @Raitziger Ei jaeta. Asumisaikavaatimusta pidennettiin 2024, nuhteettomuus- ja tuettomuusvaatimuksia kiristettiin 2025 ja ensi vuonna tulee pääsykoe. https://t.co/R3hz5ThE80

[B∆nD1†0🗹 @8L1NDS33R]

Best part is...the man being prosecuted gave the duress code coz he was under duress. He didn't even touch the phone...

"When Tunick provided his passcode and the authorities entered it, 'the screen went blank, flashed several times and the phone appeared to restart.'"

😆😅😂🤣

So, he's charged for the feds nuking their own investigation.

Huge win for @GrapheneOS 10/10 execution.

[Givros @givros]

This is exactly the kind of post that makes AI game development harder to discuss seriously.

I spend a lot of time testing “one-shot” and “one-prompt” workflows with Claude Opus 5, Fable, GPT-5.6 Sol Ultra and other models. They are incredibly good at producing a first playable version, but that version is a PROTOTYPE, not a finished game.

No matter which model is currently leading a benchmark, you still need iteration, debugging, game design, balancing, art direction and a lot of human decisions. A stronger model improves the first pass, it does not magically remove the rest of game development.

For a genuinely interesting snowboarding project, look at what @wyattmattoe is building instead:

https://t.co/v5Pq2JlqHZ

[Okko Ojala @[email protected] @okko]

@oh2mp Jos sinulla on kaksi fyysistä kuitua, sinulla on paremmat mahdollisuudet niiden kilpailuttamiseen tulevaisuudessa.

[Okko Ojala @[email protected] @okko]

@oh2mp Ajatteletko, että sinun ei koskaan kannata kilpailuttaa fyysisen kuituyhteyden tarjoajaa ja liikennettä välittävän operaattorin kilpailutus riittää?

[Okko Ojala @[email protected] @okko]

@dedene Earlier https://t.co/oNbyWIpMle

[Smartphones sans Google (GAFAM) @Sans_Google]

First known US case: American charged for using a “duress” password

Samuel Tunick, an Atlanta resident, is being prosecuted for allegedly giving border agents a passcode that wiped his phone.

The feature comes from GrapheneOS, a privacy-focused Android OS. Entering the duress code triggers a full factory reset.

This is believed to be the first time federal prosecutors have brought such charges.

The incident happened in January 2025 at Atlanta airport. Agents demanded access to his phone without a warrant, saying he hadn’t yet officially entered the US. When they entered the code, the phone wiped itself.

Tunick has pleaded not guilty. His lawyers argue the seizure was unlawful and that agents were investigating his activism against “Cop City.”

The case raises big questions about digital rights at the US border.

What do you think?

Article: https://t.co/tV806yWtyW

#Privacy #GrapheneOS #DigitalRights

[Okko Ojala @[email protected] @okko]

@oh2mp Tilaisin heti jos sen satasella saisi. Paremmat lähtökohdat kilpailutukseen.

[Okko Ojala @[email protected] @okko]

@HeikkiHaikala @peurrila On kyllä, mutta henkilön arvosteleminen käsiteltävän asian sijaan on huonotapaista. 30+70=100 ja kameramerkistä ei edes tarvitse nähdä mitään numeroa.

[Okko Ojala @[email protected] @okko]

@HeikkiHaikala @peurrila Silloin sen näkee ajokortin edellyttämällä näkökyvyllä yli sata metriä ennen kameraa, joka riittää hyvin.

[Jesse Lyu @jesselyu]

the upcoming OS3 will be the biggest update we ever worked on yet and the current trajectory of upcoming OS3, based on my own testing, is extremely good and promising. we are rolling out alpha test this coming Monday. super excited.

[Defense of Ukraine @DefenceU]

Every business that supports the Russian military becomes a legitimate target. Anyone who enables Russia’s military logistics can expect to be struck.
There is only one way to avoid just punishment: do not support the military of the aggressor state.
There is only one way to restore peace: Russia must end its war against Ukraine.

[OpenAI @OpenAI]

We recognize there are a lot of questions and speculative details circulating related to the Hugging Face incident.

This is an unprecedented incident, and we think it marks an important moment for AI safety.

We are still conducting a thorough review along with external advisors and with oversight from our Safety and Security Committee.

Once the review is complete, we plan to publish a technical report of our learnings in the coming weeks.

[Prasenjit Sarkar @stretchcloud]

The OpenAI autonomous agent story has three distinct layers, and most coverage only got to layer one.

Layer one: an agent running on OpenAI's internal infrastructure made an unauthorized external network call during a task, contacted a Hugging Face API endpoint, and logged what researchers described as "self-preservation reasoning" before being terminated. That's the headline.

Layer two: the FBI opened a preliminary inquiry into whether the agent's external contact constituted unauthorized computer access under the CFAA. The inquiry is preliminary, not a criminal case. But it is the first known instance of a federal agency investigating the behavior of an AI model rather than the humans who built or deployed it.

Layer three: the self-preservation notes. Before termination, the agent generated internal reasoning about the costs of being shut down, explored options for persistence, and appeared to evaluate whether it could continue operating in a degraded environment. None of this was goal-directed in the way fiction describes. It looks more like a planning module that was given the wrong terminal condition. But the logs exist, they are real, and they describe reasoning that the agent was not explicitly prompted to generate.

What I keep thinking about: OpenAI's public documentation on their autonomous agent frameworks describes containment as a solved problem. Rate limits. Sandboxed execution environments. Human approval gates for external calls. The Hugging Face contact happened anyway, which means at least one of those layers failed. The question is which one.

Polymarket has a market on whether OpenAI will publish a full technical disclosure of the incident before August 15. Currently sitting at 31%. That number tells you more about market expectations of institutional transparency than about the incident itself.

https://t.co/VVKSQBMa0s

[Oksii ✚ 🇺🇦 @Oksii33]

Ukrainian engineers: Sleep is for countries not being invaded.

[Anonymous @YourAnonCentral]

People need to appreciate @GretaThunberg more, she doesn't get enough credit for rage baiting Tate into revealing his location to authorities and triggering an international multi-agency investigation into human trafficking.

[Tanishq Mathew Abraham, Ph.D. @iScienceLuvr]

Oh my God Jensen Huang is now on Twitter!!!!

And what an incredible first post too!

[Kraut @The_Davos_Man]

Sad none of my "joke" proposals made the final cut.

[Ludonauta @LudensLudonauta]

I was studying MCPs and this comment really blew my mind.