BD Wang

Physical Ai

[BD Wang @barrondoira]

@thebluffingboss 她们假装很名媛，做事太low了，不体面

[BD Wang @barrondoira]

@CoCoYurealtors 其实尔湾很多德州私局，甚至很多vegas私局，都是几个创始股东，常驻reg要给host股份，相互有股份，联合起来打外人，只是布兰妮将这个放大了，德州要赢钱，就是挑比大鱼的打！

[BD Wang @barrondoira]

@CryptoApprenti1 布兰妮和peter曾经不是盟友吗？到处出双入对，兄弟助理，推测一起输了很多钱，闹掰了。但是也不该反向联合外人收割对方，有点不地道；Nik Airball要小心，下一个是你

[BD Wang @barrondoira]

@CryptoApprenti1 @TheRyanFeldman It’s very likely that she controls the lineup and gets a profit split with Ryan team.

[BD Wang @barrondoira]

@GmanPoker What’s most alarming about the cross-staking exposed now is that out of nine players at the table, almost everyone belongs to Team B or Team R. They soft-play and feed each other action, while the remaining two are treated as marks to be hunted down—it just looks incredibly dirty

[BD Wang @barrondoira]

@CryptoApprenti1 What’s most alarming about the cross-staking exposed now is that out of nine players at the table, almost everyone belongs to Team B or Team R. They soft-play and feed each other action, while the remaining two are treated as marks to be hunted down—it just looks incredibly dirty

[BD Wang @barrondoira]

@CryptoApprenti1 教练怎么跟Rayn有矛盾？教练为百万赛贡献那么大底池和流��，现在不打了，用完就放弃？只跪舔Keating？

[BD Wang @barrondoira]

Maui Sunset

[BD Wang @barrondoira]

aloha

[BD Wang @barrondoira]

Selling tools? You’re racing the model. If the LLM hits the threshold, your tool gets "compressed." 🏃‍��️💨
Selling outcomes? You’re riding the wave. Every upgrade cuts delivery costs while price holds. Margins expand; moats deepen. 📈🌊
Tools are replaced. Outcomes are compounded.

[BD Wang @barrondoira]

@Laronce__ Spot on. Detection is passive; enforcement is active. Without granular scope and real-time intervention, we're essentially handing out root access to a black box. Auditable tool calls are the only way forward.

[BD Wang @barrondoira]

A common mistake in LLM agents:

Tools are written for humans, not agents.

Example:

Click an element on the page.

Better:

Click an element.

Do not use when:
- Entering text (use fill)

Tool descriptions aren’t documentation.

They’re decision interfaces for the model.

[BD Wang @barrondoira]

Score breakdown:

Security: 10.0/10 — zero real findings
Descriptions: 3.0/10 — the real problem
Metadata: 9.2/10 — healthy project

The surprise: my automated scan initially flagged 18 issues and gave it an F.

After manual review? All 18 were false positives.

#AISecurity #MCP

[BD Wang @barrondoira]

I scanned all 3,566 source files in OpenClaw 2026.3.1 — the most popular AI agent framework (297K+ stars).

SpiderScore: 7.3/10 (Grade B)

Security is clean. But 202 tool descriptions are a mess.

Here's the full breakdown ↓
#OpenClaw #AI Security#MCP

[BD Wang @barrondoira]

Laguna Beach Sunset

[BD Wang @barrondoira]

Henry’s cuisine

[BD Wang @barrondoira]

LACMA

[BD Wang @barrondoira]

Griffith Park