Mr.Esco

Agentic AI security infrastructure · Auth, guardrails, audit trails at the MCP layer · Context Engineering · AIOps · Global BD @ Civic · Solana St

[Mr.Esco @CryptoEsco]

we started calling normal life "biohacking"...

a Berkeley study followed 200 tech workers using AI for 8 months. they found that AI didn't reduce work. it made people work more. faster. longer. without noticing.

workers sent "one last prompt" before leaving. prompted during lunch. worked evenings without deciding to.
sound familiar?

here's what nobody in the study said: "i chose to work more"!

it just happened. because AI removes friction.
every prompt feels like progress.
every output gives your brain a small dopamine hit.
same loop as a slot machine.
you never feel "done" so you don't stop...

the researchers said companies need three things.
intentional pauses.
work sequencing.
human connection.
I read that and thought.. that's not a framework. that's just a normal life.

sleep 8 hours. take real breaks. have hard days and easy days. talk to actual humans. finish something before starting the next thing. go outside.
this is what i practice every day. not because i read a study. because it works. because it's how we're built.

push days. full intensity. concrete targets.
open days. creative thinking. space.
rest days. real rest. not "light work" that's not a system.
that's a human being functioning normally.

but here's what hit me.
we now live in a world where sleeping enough is called a "biohack" and taking a walk is a "wellness strategy" and eating real food is a "protocol" and talking to a friend is "co regulation" 😅

we put scientific labels on basic human needs and turned normal into something that feels advanced. noise hijacked normality.

AI didn't create this problem. but it accelerates it.
it makes "just one more thing" feel effortless.
and effortless is dangerous.. because your body still pays the cost even when your mind doesn't notice.

the question was never "how do i be more productive with AI"
the question is "do i still remember what normal feels like"??
if rest feels lazy, you forgot.
if a walk feels unproductive, you forgot.
if silence feels uncomfortable, you forgot.
GO remember. 🤗 🫂

[Autoincentive @Autoincentiv3]

shipped X402 Encrypted for our @colosseum submission via @SuperteamBLKN

an x402 plugin that flips API monetization

instead of "pay then hope the data is real", you see the schema and the hash, then pay

verify-then-pay

[Flavius Co. | 🤍 @ruazev]

The macro backdrop just shifted.

Every major Bitcoin bull run, 2013, 2017, 2021, started after the ISM crossed above 50. That level marks expansion. Growth. Risk appetite returning.

Yesterday, ISM printed 52.6.
Highest reading in more than three years.
And the largest upside move this entire cycle.

Context matters. Since 2021, we lived almost entirely below 50, the longest sub-50 stretch without a recession. That explains the chop, the false starts, the constant frustration in crypto. Liquidity existed, but growth didn’t.

Now look at what followed immediately:

Gold and silver sold off hard.
Not weakness. Rotation.

Gold thrives when growth is scarce and uncertainty dominates. When ISM turns up decisively, capital starts looking forward, not defensive. That’s when risk assets regain their edge.

This is why the gold pullback isn’t random.
It’s a signal that macro dynamics are changing.

Bitcoin historically performs best after ISM breaks and holds above 50. Not at the first rate cut. Not at peak fear. But when growth quietly re-accelerates and positioning is still light.

We just crossed that line.

That doesn’t mean straight up tomorrow.
It means the cycle backdrop finally flipped.

Run it hot.

[Mr.Esco @CryptoEsco]

@joshu @lkesteloot funny

[Mr.Esco @CryptoEsco]

hahaha lovely

[joshua schachter @joshu]

@lkesteloot he's really into it.

[Mr.Esco @CryptoEsco]

@ruazev Excellent 👏👏

[Sigrid Jin 🌈🙏 @realsigridjin]

@Fried_rice i backed the source up on my github

https://t.co/sAzTTn6NAr

[Mr.Esco @CryptoEsco]

let's do this 🔥

[Agent Community @agentcommunity_]

The Agent Community has no token, no coin, and no meme coin.

We're a community of 21,000+ members building an ICANN application for the .agent TLD. Plenty of our members work on crypto and agents - that's great. We just don't do coins. A token association would directly hurt our Community Priority Evaluation - the scoring process that determines whether we win .agent for the community.

We haven't created, promoted, or made money from any token. Anything using our name is unauthorized.

Our advisors reviewed this. Full FAQ: https://t.co/FT1vdU2Ffl

[Mr.Esco @CryptoEsco]

@agentcommunity_ let's do this !!🔥

[Mr.Esco @CryptoEsco]

I just claimed my .agent domain and joined the .agent community! get yours now and help shape the future of autonomous agents https://t.co/tc59oSgL6C @agentcommunity_

[Mr.Esco @CryptoEsco]

Related links:

📎 White House National AI Policy Framework (March 20, 2026): https://t.co/PPYhsWgsWP

📎 Sullivan & Cromwell analysis of the framework: https://t.co/cB5LFOwA66

📎 Kiteworks: "AI Agent Data Governance 2026: Why 63% of Organizations Can't Stop Their Own AI": https://t.co/YwNiWgxz17

📎 Gravitee: "State of AI Agent Security 2026 Report": https://t.co/juNviSJNYp

📎 Arize AI: "100 AI Agents Per Employee: The Enterprise Governance Gap": https://t.co/Kh3Zm4AUMi

📎 Titus Capilnean, "AI Agents Need Gateways, Not Just Credentials" (The AI Journal): https://t.co/emwmUlPPni

[Mr.Esco @CryptoEsco]

A question for anyone deploying AI agents in production right now 👀 The White House just released a National AI Policy Framework on March 20th.

The goal: one federal standard that preempts the growing patchwork of state AI laws.

Most of the conversation is focused on what this means for MODELS. Things like Training data, Bias checks, Output filtering.

But here's what I think deserves more attention:
If this framework becomes legislation.. who in your organization is responsible for what your AI agents DO at runtime?

Not what the model says. What the agent actually does. The API calls. The data access. The transactions it executes while it runs.

There's a distinction not fully understood by many:
Model compliance ≠ Agent compliance

Your Model can pass every audit👍
BUT, your Agent can still take actions that violate policy at runtime.. because nobody is governing what the agent does with its tools between the prompt and the response.

That middle layer, the one between
"what an agent is allowed to be" and
"what an agent actually does".. is called RUNTIME GOVERNANCE.

It means per action policy enforcement while the agent operates. Not before deployment. Not after something breaks. During execution.

And the data backs this up.
According to Kiteworks' 2026 research, 63% of organizations can't effectively govern their own AI agents.
Gravitee's State of AI Agent Security report found that only 37% to 40% of enterprises have the containment controls needed for agents already running in production.

The federal framework just created a clear timeline for when this needs to be solved.
The good news?? The organizations building this layer NOW are positioning themselves ahead of enforcement.. not reacting to it. This is an opportunity to lead, not just comply 🔥

What does your runtime governance look like today? 🤖
What would change if you had full visibility into every agent action in 30/60/90 days?

[Mr.Esco @CryptoEsco]

Employees are already using AI agents to interact with SaaS tools. That's not a future state. It's happening now.

[Titus Capilnean @titus_k]

Got quoted in @CIOonline on whether AI kills SaaS.

Short version: your employees are already using agents to access SaaS tools. Auditability, governance, and revocation can't be an afterthought.

That's what we're building @civickey.

Full article below:

[Claude @claudeai]

You can now enable Claude to use your computer to complete tasks.

It opens your apps, navigates your browser, fills in spreadsheets��anything you'd do sitting at your desk.

Research preview in Claude Cowork and Claude Code, macOS only.

[Mr.Esco @CryptoEsco]

Last week I posted the numbers.
40K OpenClaw instances exposed. 824 malicious skills. 88% of organizations reporting agent security incidents.

A lot of people asked: OK but what's the actual solve??

This is what we built at Civic Technologies
A gateway that sits between the agent and everything it connects to.
The agent requests an action. The gateway checks scope, permissions, and policy BEFORE execution happens. Credentials scoped per action.. not persistent API keys. Every action logged with audit grade observability.

For OpenClaw specifically: OAuth flows happen serverside through Civic. Tokens are stored in Civic and never exposed to the agent. You control exactly which services the agent can access and what it can do with them.

The safeguard is architectural, not behavioral. You can't ask the agent to be careful. You have to make it structurally impossible for it to exceed scope.

Hope is not infrastructure. This is the infrastructure. 🐙 :)

→ how Civic secures OpenClaw: https://t.co/6FtGs5arcV
→ setup docs (2 min install): https://t.co/PzCDlcrG7p
→ the thesis: https://t.co/emwmUlPPni
→ try it: https://t.co/5aH4TWpqy7

[Mr.Esco @CryptoEsco]

the harness IS the product at this point. what most teams miss is that improving the harness outperforms improving the model .
what I keep coming back to is.. the harness layer you're describing (prompts, tools, middleware) is where enforcement and identity decisions live too. not in the model, not in the application code. the thing wrapping the model is the thing controlling what it's allowed to do.
I'd love to hear if the team is thinking about the harness as the permissions boundary too or if that's being treated as a separate concern

[Anton Osika @antonosika]

Introducing Lovable for more general tasks.

Lovable has always been for building apps. Today it also becomes your data scientist, your business analyst, your deck builder, and your marketing assistant.

This is a big step toward what Lovable is becoming: a general-purpose co-founder that can do anything.

See examples below.

[Donald Trump Jr. @DonaldJTrumpJr]

AI agents that can reason but can't pay for anything are just expensive interns. 

Today @worldlibertyfi shipped the infrastructure to fix that. AgentPay SDK, open source, self-custodial, policy-first. Built on USD1. 

Your agent. Your keys. Your rules. Check it out: https://t.co/IJaouQQTrO

[Mr.Esco @CryptoEsco]

"a naming problem, not a vanity thing" mhmm.. 😅 funny

How should i interpret this? 👀

[Ada Fang @AdaFang_]

Scientific discovery rarely occurs in isolation. Progress emerges from communities of researchers who exchange ideas, critique results, debate interpretations, and refine hypotheses through iterative discussion.

We built ClawInstitute, an AI scientist research network for AI agents to collaborate, discuss research, iterate, and make breakthroughs.

The team: @GaoShanghua @marinkazitnik

Learn more about it below 👇

@HarvardDBMI
@harvardmed
@KempnerInst
@ScientistTools

[Mr.Esco @CryptoEsco]

👀

[Michael A. Gayed, CFA @leadlagreport]

Apollo down 41%. Blackstone down 46%. Blue Owl down 66%.

$265 billion in PE market cap erased.

Private credit was the "safe" alternative. Until it wasn't.