Raf

Engineering Decentralized 🌎

[Raf @proofofyoon]

Reduce Dependency for Freedom

[Teddy - PolyBackTest.com @Bitcoin_Teddy]

Telegram CEO says, “I’ve had a few hundred million dollars in my bank account or in Bitcoin since 10 years ago.” 🤯

“My no. 1 priority in life is my freedom.”

[Raf @proofofyoon]

@KKDW_KOREA 아.. 규제를 완화하기 위한 장치로 보고 계시군요. 일리가 있는 말인거 같네요. 그런 의미라면 지금 이상으로 덩치를 키울 필요는 없을것 같기도하네요

[Raf @proofofyoon]

@seti_park 아하..! MAC 으로의 우회가 직접 CNN 으로 주입하는 방식으로 되고 비트 최적화까지 이어진거군요. 감사합니다!

[Raf @proofofyoon]

Just Use DeFi

[퓨처워커 미국증시 리서치 @Future__Walker]

미래에셋증권이 정부의 정책에 따라 고객들의 해외주식을 강제로 매각할 수 있다고 공지해 논란이 일고 있다.

[Raf @proofofyoon]

커서팀이 자체 에이전트로 일주일간 300만줄 이상의 코드를 작성하면서 느낀 점

- 복잡한 계층 구조와 세분화된 역할 분담은 대부분 병목과 성능 저하를 가져옴
- 최소한의 구조와 자율성이 가장 안정적이고 높은 품질의 결과로 이어짐
- 이는 인간 조직에서 오랫동안 관찰되어온 패턴과 유사

[Lee Robinson @leerob]

Some cool things from our research on long-running agents:

1. We built a custom harness around GPT-5.2 so that any task could spawn thousands of concurrent agents.
2. Even with a lot of tokens and $$$ spent on this research, the results are still far from "production ready" by any means. It's an interesting look at what the future might be like, but we have a long ways to go.
3. The model and harness mattered a lot, but a surprising amount of tuning came from small prompt tweaks and experiments! For something more real, this is where having good evals would really help.
4. The best solution ended up being more simple than we expected. We tried to borrow ideas from distributed systems (think: Kubernetes for agents) or how you'd structure real teams. But these often made things worse. Agent design requires new patterns.
5. This seems now widely accepted, but I was surprised that GPT-5.2 was so much better for long-running tasks. My first experiments with Pixo (my Rust library) were all with GPT-5.1 Codex using the alpha version of this tool (lovingly called "grind mode" internally).

P.S. We also fixed some of the compilation issues on the browser repo and got CI running. The README has some nice architecture diagrams as well.

Blog: https://t.co/wP5DtgADuy
Code: https://t.co/128euoD0WG

[Raf @proofofyoon]

@affaan Great article for reshaping my claude-code setups :) One question is "is there any you own tips for optimizing context window management?" Thank you!

[Raf @proofofyoon]

@r2Jamong 그렇군요. 그럼 솔라나는 이런 상황을 어떻게 보고 있나요? 뭔가 중앙화 이슈는 있지만 오히려 성능 좋은 밸리데이터들이 남게되어 질적으로 좋아지려나요

[Raf @proofofyoon]

@USAnt_IDEA 그러게요 ㅋㅋ 잭도시는 앱을 만들면서 이런 일을 예상했을까요?

[Raf @proofofyoon]

@talkevergreen 앗 그럼 코인베이스가 usdt 를 페어를 리스팅할 가능성이 있을까요?

[김민전 @kimmeengeon]

공유합니다.

우리나라에서 머스크처럼 얘기하면
“음모론자”가 되는데ㅠㅠ

[Raf @proofofyoon]

기존 LLM: "한번 학습된 가중치 = 영원“ > 긴 문맥 저장필요 > 메모리 수요 증가
TTT: 컨텍스트 기반으로 순간순간마다 조금씩 학습 > 저장 안함

아직 초기 단계지만 상용화되면 HBM 수요에 영향을 끼칠수도?

[NVIDIA AI Developer @NVIDIAAIDev]

We are entering a new era for LLM memory. 🧠

In our latest research, End-to-End Test-Time Training, LLMs keep learning at test time via next-token prediction on the context – compressing what they read directly into their weights.

Learn more: https://t.co/N71zsmC2ay

[Raf @proofofyoon]

진심으로 부럽다.. 이런 법안을 낼 수 있는 미국도 대단하고

[Congressman Randy Fine @RepFine]

Huge News! Today, I am proud to introduce the Greenland Annexation and Statehood Act, a bill that allows the President to find the means necessary to bring Greenland into the Union. 

Let me be clear, our adversaries are trying to establish a foothold in the Arctic, and we can’t let that happen.  

By acquiring Greenland, we would prevent our adversaries from controlling the Arctic Region and secure our northern flank from Russia and China.

[Raf @proofofyoon]

@damnang2 이런 구조니 인재 밀도가 높아질 수 밖에 없을거 같아요. 재밌게 잘 읽었습니다!

[Raf @proofofyoon]

👆

[Anakin @aegis721]

원화가치가 떨어지니 이런일이 벌어지네요

[Raf @proofofyoon]

부럽다

[주식열차 @stocktrainkorea]

그린란드 담판 예정
주민들이 병합동의하면
미국시민권과 1인당 10만달러 추진 중 ㅎㄷㄷ

[Raf @proofofyoon]

@Simba_Doges 확실히 패시브 인컴을 가져다줄 수 있다는 점에서 기관들 입장에서 매력적인거같아요