sleepybear

📚 🦔🎡

[Machina @EXM7777]

https://t.co/2cnMTXTMl1

[Anthropic @AnthropicAI]

We're publishing our most detailed threat intelligence report to date.

It covers how people tried to misuse Claude—for cyberattacks, influence operations, surveillance, biology, and building weapons—and how we found and stopped them.

We disrupted every operation in the report, and used the lessons from them to strengthen our safeguards. Where appropriate, we also shared what we found with authorities and other AI companies.

These cases are not typical: we’re highlighting some of the most sophisticated misuse we’ve seen. But they’re especially important to discuss, because they show us where AI misuse is headed, where our safeguards work, and where they need to improve.

We’re publishing this report so others can spot the same activity on their own platforms, and so we can give the public a clearer view of how emerging threats develop.

Read the report: https://t.co/0EJUnYEgfz

[Anmol K Sachan @FR13ND0x7F]

I built a Burp Suite extension for automated LLM prompt injection testing.

Because manually pasting jailbreaks into Repeater is not a methodology.

🤖 LLM Injector - thread

[1/6]

[sleepybear @TSripo]

@FR13ND0x7F thx for your burp suite extension a lot, it really help with my internal testing very much

[Rohan Paul @rohanpaul_ai]

This is outside Anthropic building in San Francisco.

you can see that emotions are all over the place when it comes to AI models and their creators.

[Anthropic @AnthropicAI]

A statement from Anthropic CEO, Dario Amodei, on our discussions with the Department of War.

https://t.co/rM77LJejuk

[Paul | BAND (🅑) 🐰 @PNattapatsiri]

GMembit,

Our next @membit_ai upgrade delivers real-time segmentation across HUNDREDS of crypto narratives and ~5× smarter context.

Thanks to:
• 60k data hunters
• 1M+ fresh posts daily
• Engineers who don’t sleep

✅ Going live in the next 2-3 days.

More to come. 🐰🐇🐇🐇

[Paul | BAND (🅑) 🐰 @PNattapatsiri]

https://t.co/wdHTNxwNLo

[Paul | BAND (🅑) 🐰 @PNattapatsiri]

12,500 Hunters.
SO. MUCH. DATA.

You guys are insane — @membit_ai community just unlocked the potential of real-time context layer for AI and it's getting real.

Collective intelligence is no longer a dream, it’s happening right here.

Our clustering is getting absolutely fried 🔥 but this massive scale of data is exactly what we need to evolve fast.

Day 3. We’re just getting started.

[Paul | BAND (🅑) 🐰 @PNattapatsiri]

GMembit!

It’s been an incredible 48 hours since Epoch 1 of @membit_ai went live. The response has been beyond anything we imagined.

A transparent look at our progress 👇

👍 The Great:
===========
- Community Growth: Over 8,000 of you joined as Data Hunters, with a ~50% conversion rate from 15,000+ website visitors. Welcome!
- Data Velocity: Together, you've already contributed nearly 500,000 unique posts, building a massive foundation of real-time context.

😅 The Growing Pains:
===========
- We Got Hit Hard: Our nascent infrastructure was slammed with over 1,000 requests per second, handling 10 million requests and counting.
- Rewards Were Too Low: Our initial point system was too conservative. We know it's been hard to hit the cap, even for the most dedicated scrollers.
- Pipeline Clogged: The sheer volume of data temporarily bottlenecked our clustering system, impacting the quality of the context.

⚡️ The Fixes (Rolling Out Now):
===========
- Hardened Infrastructure: We've significantly upgraded our backend. Handling this scale is a familiar challenge at @bandprotocol, and we're now ready for more.
- Better Rewards: A more generous point distribution is being gradually rolled out to properly reward your time and curiosity.
- Smarter Clustering: We're upgrading our data engine to process millions of posts in real-time, ensuring the context layer stays fresh and high-quality.

Remember, this is just Day 2. We're building this ground-breaking context layer for AI agents in public, with you.

What an exciting time.

[sleepybear @TSripo]

"Maker" & "Manager" schedule
- Manager: divided into a hour-block of meeting
- Maker: creator need long, uninterrupted period of time
The problem happen when 2 schedule collide
For startup in Maker phrase, Manager shouldn't meet during Maker time
ref: https://t.co/vMylicMShZ

[Paul | BAND (🅑) 🐰 @PNattapatsiri]

https://t.co/ovzsvwj532

[sleepybear @TSripo]

if board shift into less range advantage for raiser
the bet size for protecting strong but vulnerable become more polarized (~1.2x)
- best hand need protect: AQ, AJ, AT
draw that add into this range
- gutshot w/o FD
- FD w blocker around 9: J8, T8, 87, 86

[sleepybear @TSripo]

[Poker][self study]
C-bet with dynamic board
starting with rainbow connected board: AKJ
most of this range bet: 1/3-1/2
one that choose to bet larger (1/2)
is strong hand that vulnerable
- top pair + top kick: AQ, AT

[sleepybear @TSripo]

3. draw hand
- gutshot & OESD
-- w/o FD: sc-bet
-- w FD: check
-- os-comb
-- w ♣️: sc-bet
-- w/o♣️: 50-50
- BDFD
-- A-high: bet both high & low kick; 9, 2
-- bottom range: 4 - 7: sc-bet
-- all other: check

4. trash hand
- play similar trend like BDFD

[sleepybear @TSripo]

draw type hand
most of them play as a check or small bet
one that alway bet small (1/4-1/3)
- FD + 2-over
- gutshot + FD

[sleepybear @TSripo]

board: QQ + K
play 2 sizes

1. best hand cat
- quad, full house, set,
- top pair
-- top kick: small c-bet
-- low kick: check

2. marginal hand
- underpair (+ AA): check
- low pair: 66-
-- mostly bet, esp. with ♣️ unblock FD i guess