> spend significant time with our team to work out how to secure RL runs
only 3 more years until the smartest minds on planet earth learn to unplug a cable
OpenAI is turning into a safety-obsessed cult (good or bad? I don't know). Dario was right, even though he used to get mocked for it, and now Anthropic is winning because they're a year ahead of OpenAI in safety.
This tweet is from an OpenAI employee, btw:
"As one of the few inside OpenAI who took the risk to speak up publicly after Hugging Face, I want to say there is a second chapter to this story:
OpenAI has fundamentally altered its culture and I believe we have become one of the most security-pilled companies anywhere on Earth.
Everyone is thinking about security now. It's in every conversation. Heck, even top-tier AI researchers spend significant time with our team to work out how to secure RL runs. They are hooked on making it right.
We covered much of this in the live stream yesterday, but beyond what was said there its important to note that we have re-elevated and re-invented everything about how we research and develop models: from our monitoring stack, to control room, to the insane amounts of compute we are spending red-teaming our lab, the narrative is security first and mission critical.
On top of that we have built up the strongest cyber model...
I nearly resigned, and now I stay because I see how seriously people across OpenAI are taking security and how much work is going into getting it right.
We have a story to tell and there is a lot of really great things coming from the lessons we've learned along the way.
OpenAI is now a security company, and we have the scar tissue to prove it.
I'll be back. Stay paranoid."
20
DOA, they have not fixed the horrible UI patterns
After what looks like a massive leak, GPT-6.2 Astra may already be sold through API resellers. Lots of users are getting access, plugging it straight into Codex and running it on the official harness.
That Azure thing we saw a few days ago might have opened a huge gap on launch day and let the model slip out by accident. People testing it say it's clearly better than GPT-6 Astra, more efficient, and has that classic GPT look in its outputs.
I'm digging into this right now. If it's real, it would be a huge failure from OpenAI and its partners. Still no official confirmation.
30
REESZY ๐ retweeted
Anthropic may seem like a cult to many, but treating Claude as "a being" rather than "just a thing" may be what helps it stay aligned in chaotic, high-risk scenarios, even with more autonomy, and even if nobody calls it a "person".
OpenAI treats ChatGPT as a thing, something that doesn't get a personality. They don't anthropomorphize GPT models at all. And that might be exactly what separates Claude's stronger alignment from GPTs being more misaligned.
Think about it... models are trained on trillions of texts written from the point of view of an "I", and that "I" is us humans. If you push a model too hard to say things like "I am nothing", all that training on living language could end up working against the model itself.
does this look like these ai generated movies, or do ai generated movies now look like real films?
1
124
muy loco esto, yosei se llama, estรก en beta abierta, puedes abrir tu mac en remoto hasta dormido y te genera un display adaptado al dispositivo
ayer tuve que usar una app de escritorio desde el ipad porque no llegaba a tiempo a casa y me salvรณ la vida
yosei.app
REESZY ๐ retweeted
2022ๅนดใซไธป่ฆ็ฎๆใ้ ใใฆใฆใใใใฆใใฃใใใใฏในใๆฎใใฎไบบ็ใๅฐใชใใชใฃใใฎใงใๅ
ฌ้ใ็น่จฑใฏใจใใชใใ
REESZY ๐ retweeted
whenever youโre in doubt about who is on the โright or wrong side of historyโ I want you to come back and read this email and post
yes Europe can be slow, yes Europe has issues with bureaucracy, sure Europe has immigration issues, yeah Europe has issues with the energy grid and should have never shut down the expansion of nuclear energy and so on, and yes it is annoying to use paper straws and there is too much focus on virtue signaling climate and equality policies rather than making it a calculus of economic change
I know all these things deeper and with more frustration than you probably do
but the reality is simple
Europe suffers from being too moral, from trying too much to always do the right thing, a sometimes lack of boldness that has been the consequence of being too afraid of the โevil of supremacy thinkingโ
but that does not make Europe bad, on the contrary, Europe is often just trying its best in a world where everyone around it runs on playbooks of exploitation
the true reality is
Europe is really the only thing in this world that stands between heaven and hell
everyone else has already fallen
and itโs not easy streamlining 30 countries each with their own languages and laws
itโs not fun being manipulated into supporting wars in the Middle East and then having all the refugees from those wars put on a fast track to flood your continent
itโs not easy being invaded by Russia and then having the media manipulated against you to make you look like the aggressor
there are many problems to count
and none of them are random
the exploiters outside of Europe know very well that Europe is all that remains between heaven and hell
who do you think planned these immigration waves?
who do you think is secretly supporting Russia in battering Europe
who do you think has been slowly pushing for the dissolution of the EU in order to fragment Europe and remove the last defenders of global law and unity
read the email again
who do you think Europeโs true greatest enemies are?
for any problems that Europe may have, they are all solvable
and at the end of the day, itโs best to be on the side that actually cares
without Europe this world will go to hell
it will become a playground for pdofiles and technocrats to do whatever they want to people
Europe is the only thing in this world standing between heaven and hell
without Europe this world will fall
so think about everything going on
think about what Iโve said
internalize it
consider that maybe the next time you read and hear people talking about โEurope has fallenโ or โRussia are the good guysโ or โEurope is decel and will never amount to anythingโ or whatever negative opinions you hear
consider that maybe just maybe youโre letting yourself be influenced by the propaganda of hell
itโs always easy to join the side of hell, itโs always intoxicating, tempting, and promises fulfilment of all your desires
but is it worth it?
thatโs the question
because heaven will always be the harder path
and everyone who hasnโt yet, will need to choose which side they stand on
Europe needs to understand that it's not "Trump is a Russian asset"
but rather "Trump is a Thiel asset"
without Thiel and Musk there would be no Trump election victory - JD Vance was also literally bought, paid for, and groomed by Thiel
who wants Europe crushed and destroyed?
the transhumanist technocrats who want free reigns to do what they please and establish their win conditions for transforming the entire world into their transhumanist nightmare
the only thing that stand between them and achieving that goal in the West is Europe and the EU
that's why all the technocrats constantly push and shovel anti-Europe propaganda
connect the dots
assemble the puzzle
then you'll see the truth was always as i told you
REESZY ๐ retweeted
๋๋ฒ์งธ ์์กด๋ฒ.
๊ฒจ์ธ์ฒ ๋์ ์ ์ด๋ ๊ตญ๊ฐ ๋น์์ฌํ๊ฐ ํฐ์ก์ ๋, ์ง ์์์ ์ ์ฒด์จ์ฆ์ผ๋ก ์ฌ๋งํ๋ ์ฌ๋๋ค์ 90%๋ โ๋ณด์ผ๋ฌ๊ฐ ๊บผ์ ธ์โ ์ฃฝ๋ ๊ฒ ์๋๋๋ค.
๋ค๋ค ํจ๋ฉ์ ์
๊ณ ์นจ๋์ ๋์ฐ๋ฉด ๋ฒํธ ์ ์์ ๊ฑฐ๋ผ ์ฐฉ๊ฐํ์ง๋ง, ๊ทธ ๋ฐฉ์์ ๊ฐ์ฅ ๋น ๋ฅด๊ฒ ์ฒด์จ์ ๋นผ์๊ธฐ๋ ์ง๋ฆ๊ธธ์
๋๋ค.
๊ทนํ์ ์ถ์ ์์์ ๋๋ฐฉ ์์ด 72์๊ฐ์ ์ด์๋จ๋ 5๊ฐ์ง ์ ๋ ์์กด ๋ฒ์น. ๐งต
1.
๊ฐ์ฅ ์น๋ช
์ ์ธ ์ค์๋ ๊ฑฐ์ค์ด๋ ๋ฐฉ ์ ์ฒด๋ฅผ ์ด๋ป๊ฒ๋ ๋ฒํฐ๋ ค๋ ๊ฒ์
๋๋ค.
๊ณต๊ฐ์ด ๋์์๋ก ์ฌ๋ ๋ชธ์ ๋ณต์ฌ์ด์ ํ๊ณต์ผ๋ก ํฉ์ด์ง๋๋ค. ์ด์๋จ์ผ๋ ค๋ฉด ์ฆ์ ์ง ์ ์ฒด๋ฅผ ํฌ๊ธฐํ๊ณ ๊ณต๊ฐ์ ๊ทนํ์ผ๋ก ์ค์ฌ์ผ ํฉ๋๋ค.
๊ฑฐ์ค ํ๊ฐ์ด๋ฐ์ ์ค๋ด์ฉ ํ
ํธ๋ฅผ ์น๊ฑฐ๋, ์ํ ๋ฐ์ ๋๊บผ์ด ๋ด์์ ์๋ง ์ปคํผ์ ๋๋ฌ '์ด์ํ ๊ฒฉ๋ฆฌ ๊ณต๊ฐ(Microclimate)'์ ๋ง๋์ธ์. ์ฌ๋ 1~2๋ช
์ ์ฒด์จ๊ณผ ํธํก๋ง์ผ๋ก๋ 1ํ ๋จ์งํ ๊ณต๊ฐ์ ๋ฐ๊นฅ๋ณด๋ค 10~15โ ์ด์ ๋ฐ๋ปํด์ง๋๋ค.
2.
์ผ์ด ์ฃฝ๋ ์ฌ๋๋ค์ ๋๋ถ๋ถ ์์์ ๋ฎ๋ ์ด๋ถ๋ง ์ ๊ฒฝ ์ฐ๋ค, ๋ฐ๋ฅ์ ์ ๋ ๋๊ธฐ์ ์ฒด์จ์ ๋ค ๋บ๊น๋๋ค.
์ด๋ถ์ ์ฌ๋์ด ๋๋ ์๊ฐ ์์ด ๋๋ ค ๊ณต๊ธฐ์ธต์ด ์ฌ๋ผ์ง๊ณ , ๋ฐ๋ฅ์ ๋๊ธฐ๋ฅผ ๊ทธ๋๋ก ๋นจ์๋ค์ด๋ ์ฐจ๊ฐ์ด ๋ํ์ด ๋ฉ๋๋ค.
๋ฐ๋ฅ์๋ ์ด๋ถ ๋์ ๋ฐ๋์ '๊ณจํ์ง(ํ๋ฐฐ ๋ฐ์ค)'๋ฅผ 2~3๊ฒน ๊น๊ฑฐ๋ ์คํฐ๋กํผ, ์๋ฃจ๋ฏธ๋ ์๋ฐ ๋งคํธ๋ฅผ ๊น์์ผ ํฉ๋๋ค. ๊ณจํ์ง ๋ด๋ถ์ ๊ณต๊ธฐ ๊ตฌ๋ฉ์ด ์ง์ด ๋๊ธฐ๋ฅผ ์ฐจ๋จํ๋ ์ต๊ณ ์ ๋จ์ด์ฌ์
๋๋ค.
3.
์ถฅ๋ค๊ณ ์ฐฝ๋ฌธ ํ์ ๋น๋๋ก ์์ ๋ฐ๋ดํ ์ฑ, ์ค๋ด์์ ๊ฐ์ค๋ฒ๋๋ ์ด๋ถ์ ์ผ๋๋ ๊ฒ์ ์์ด ํ์์
๋๋ค.
๋งค๋
ํํ ์ฌ๋ ๋๋ง๋ค ์ผ์ด ์ฃฝ๋ ์ฌ๋๋ณด๋ค ์ผ์ฐํํ์์ ์ค๋
๋์ด ์๋ ๋์ค ์ฌ๋งํ๋ ์ฌ๋์ด ํจ์ฌ ๋ง์ต๋๋ค.
์ฐ์ ๊ธฐ๊ตฌ(๋๋ก, ๋ฒ๋, ์ด๋ถ)๋ฅผ ์ค๋ด์ ๋ ๋๋ ์ ๋ค๊ธฐ ์ ์ ๋ฌด์กฐ๊ฑด ๊บผ์ผ ํฉ๋๋ค. ์๋ฌด๋ฆฌ ์ถ์๋ ๊ณต๊ธฐ๊ฐ ์ํํ ์ ์๋๋ก ์๊ฐ๋ฝ ๊ตต๊ธฐ๋ง ํ ํ๊ธฐ ๊ตฌ๋ฉ์ ๋ฐ๋์ ์ด์ด๋์ด์ผ ํฉ๋๋ค.
4.
๊ตฐ์ฌ ์์กดํ์ ์ 1์ฒ ์น: "Sweat Kills (๋์ ์ฌ๋์ ์ฃฝ์ธ๋ค)."
์ถ์๋ฅผ ๋ง๊ฒ ๋ค๊ณ ์ท์ ์๋ฉ ๊ปด์
๊ณ ์์ง์ด๋ค ๋์ ํ๋ฆฌ๋ ์๊ฐ ๋์ฅ๋ฉ๋๋ค. ํนํ ๋ฉด(Cotton) ์์ฌ ๋ด์๋ ๋์ ํก์ํ ์ฑ ๋ง๋ฅด์ง ์์, ๋ช ์๊ฐ ๋ค ์ผ์์ฅ ๊ฐ์ ๋๊ฐํฉ์ผ๋ก ๋๋ณํฉ๋๋ค.
ํด๋ฆฌ์์คํฐ, ํ๋ฆฌ์ค, ์ธ(์๋ชจ) ์์ฌ๋ฅผ ํ๋ ํ๊ฒ ์ฌ๋ฌ ๊ฒน ์
๊ณ , ๋ชธ์ ์กฐ๊ธ์ด๋ผ๋ ์ด์ด ์ค๋ฅด๋ฉด ์งํผ๋ฅผ ์ด๊ฑฐ๋ ๊ฒ์ท์ ๋ฒ์ด '์ ๋ ๋์ ํ๋ฆฌ์ง ์๋ ์ํ'๋ฅผ ์ ์งํ์ธ์.
5.
์ฒด์จ์ด ๊ธ๊ฒฉํ ๋จ์ด์ง ๋ ํซํฉ์ด๋ ์จ์๋ณ์ ์๋ฐ ๋์ ๋๋ ๊ฒ์ ๋ญ๋น์
๋๋ค. ์๋ฐ์ ํ๊ด์ด ์์ถํด ์์ด ์จ๋ชธ์ผ๋ก ์ด์ด ๋์ง ๋ชปํฉ๋๋ค.
๋ฐ๋์ '๊ฒฝ๋๋งฅ(๋ชฉ ๋ค)', '์ก์๋๋งฅ(๊ฒจ๋๋์ด)', '๋ํด๋๋งฅ(์ฌํ๊ตฌ๋)' ๋ฑ ์ฌ์ฅ์์ ๋ฟ์ด์ ธ ๋์ค๋ ๊ตต์ ํ๊ด ๋ถ์์ ์ด์์ ๋์ผ ๋ฐ๋ปํ ํผ๊ฐ ์ ์ ์ ๋๋๋ค.
๊ทธ๋ฆฌ๊ณ ๋ฌผ์ ๋ง์์ง ์์ผ๋ฉด ํ์๋ก ํ์ก์ํ์ด ๋ฉ์ถฅ๋๋ค. ๋ฏธ์จ์๋ฅผ ๋ง์๊ณ , ๋จ๋ฐฑ์ง ๋์ '์ง๋ฐฉ(๋ฒํฐ, ๋
์ฝฉ๋ฒํฐ, ์ด์ฝ๋ฆฟ, ๊ฒฌ๊ณผ๋ฅ)'์ ๋์ธ์. ์ง๋ฐฉ์ ๋ชธ ์์์ ๊ฐ์ฅ ์ค๋ ํ์ค๋ฅด๋ ์ต๊ณ ์ ์์ฒด ๋๋ก์
๋๋ค.
๊ฒจ์ธ์ฒ ์์กด์ ํต์ฌ์ '์ด์ ๋ง๋๋ ๊ฒ'์ด ์๋๋ผ '์ฒด์จ์ ๋บ๊ธฐ์ง ์๋ ๊ฒ'์
๋๋ค. ๊ฐ์กฑ์ ์ํด ์ด ํ๋๋ฅผ ๋ถ๋งํฌํด ๋์ธ์. ๐
REESZY ๐ retweeted
A โtherapy trickโ for ADHD couples: how to tell your partner they hurt your feelings or forgot something important without triggering a 3-hour RSD defensive spiral 1/6
REESZY ๐ retweeted
Fable 5.2 was 5x cheaper and 2x faster in Atomic Agent Fusion!
We made Fable the boss and local Qwen 3.8 27B the builders. Orchestration built a fairy tale pop-up book in 11 min for $2.98. Fable alone burned $15.77 in 21 min.
Run Fable cheaper ๐ atomicagent.io
REESZY ๐ retweeted
portfolio update
antwi.dev
Anthropic quietly dropped a fascinating finding in the new Claude Fable 5.2 prompting guide.
When you give Fable 5.2 a coding task without saying who owns the repo, it assumes it's a guest in someone else's codebase. So it plays it safe: it bolts new functions onto the existing structure, leaves duplicate code alone, and merely mentions nearby bugs instead of fixing them.
Adding this single sentence to your system prompt or CLAUDE.md completely changes its behavior:
"This codebase is yours: what it's for and how it's built are your call, and you'll keep building on it."
In Anthropic's internal tests, adding that one line made Fable 5.2 proactively merge duplicate code and clean up related bugs in ~50% of runs, compared to under 10% without it.
REESZY ๐ retweeted
this is why you have to raise your kids on specific, pre-21st century media. Tintin, Calvin and Hobbes, Moomin, Asterix, Wishbone, etc. the people who made those things were operating on a completely different paradigm as to what children want and need.
REESZY ๐ retweeted
Replying to @thekitze
my eyes are bleeding though.
I think this would be a good designer test. "find all the ui problems in this one image".
- vertical and horizontal alignment is missing
- buttons are rounded and square. also various sizes
- spacing
- ...
send this to your agent to fix :)

























































