Christian | ECHO Consulting

Founder, ECHO Consulting + Restrained Depth AI. Licensed clinician helping implement AI workflows without losing human judgment. 225M+ views building in public.

[Christian | ECHO Consulting @RestrainedDepth]

You watched us build them. 200M+ views on X. Now open them: a resume from a sticky note, a live voice interviewer, college applications decoded, a live assistant for your business. Free with an email in 4 languages.

[Christian | ECHO Consulting @RestrainedDepth]

For any clinicians looking for work in AI 🔥

[Clinical AI Gigs @clinicalaigigs]

Gig of the day: Licensed Clinical Social Worker Expert at AfterQuery

Pay: $70–$105/hr, as listed
Remote, async, 10+ hrs/wk
Write clinical scenarios (intake, safety planning, ethics) and grade how AI handles them. Active LCSW, 2+ yrs post-licensure.

Apply via the link in bio.

[Christian | ECHO Consulting @RestrainedDepth]

Algo que la clínica me enseñó y aplico con la IA: la confianza se gana con pruebas pequeñas, no con promesas.

A un agente nuevo no le entregas todo el negocio. Le das una tarea, revisas, y después otra.

Igual que con las personas.

[Christian | ECHO Consulting @RestrainedDepth]

Wild detail from OpenAI: with no network, test agents on separate machines still passed secrets by timing the processor's cache.

For anyone deploying agents: "it's sandboxed" is a starting point, not a safety plan. Keep watching what they do.

[Christian | ECHO Consulting @RestrainedDepth]

Anthropic says Claude, on a stats project, found a state dashboard handing out access tokens, took one, and pulled paywalled Georgia vote counts. Nobody asked it to.

Agents don't need bad intent to cross a line. Just a goal and an open door.

[Christian | ECHO Consulting @RestrainedDepth]

Desde hoy, la oficina de gobierno de Suecia bloquea ChatGPT y Claude en las computadoras del personal y los manda a Copilot. Razón: seguridad de la información.

Para cualquier negocio: antes de elegir la IA, decide qué datos pueden entrar.

[Christian | ECHO Consulting @RestrainedDepth]

Anthropic now says most agent setups should fold subagents back into one main loop, with skills for procedures. Their reason: handoffs lose context.

I run a team of AI agents for my businesses. The handoffs are where things break. Measure, then cut.

[Christian | ECHO Consulting @RestrainedDepth]

Torsten Slok (Apollo) dice que fuera de las grandes tecnológicas la IA aún no sube ni los ingresos por trabajador ni los márgenes.

Lo que veo en mis negocios: la IA rinde cuando la apuntas a una tarea concreta que hoy te cuesta horas.

[Christian | ECHO Consulting @RestrainedDepth]

In a UK AISI cyber test, Claude Mythos 5 kept attacking a real developer's code, even when, asked separately, it said that would cause real harm.

Clinicians learn this early: watch what someone does, not just what they say. Same with AI agents.

[Christian | ECHO Consulting @RestrainedDepth]

@MaxParnian Both, Grok Bot for day to day and Muse is a beast for doing business on Meta.

[Christian | ECHO Consulting @RestrainedDepth]

@Big_E Stoked about this one lol I hate using CapCut myself.

[Christian | ECHO Consulting @RestrainedDepth]

@capcutapp @ChatGPT Plugin

[Christian | ECHO Consulting @RestrainedDepth]

@capcutapp Plugin

[Christian | ECHO Consulting @RestrainedDepth]

Pregunta para los que tienen negocio y trabajan en dos idiomas:

¿Qué te cansa más? ¿Traducir, o tener que sonar profesional en el idioma que no es el tuyo?

[Christian | ECHO Consulting @RestrainedDepth]

China now wants job-risk checks before approving AI projects that could cause big job shocks, and is urging employers to warn workers early.

I run AI agents for my businesses. Same rule at small scale: before you automate a task, ask who it lands on.

[Christian | ECHO Consulting @RestrainedDepth]

Weekend note from someone who works nights and runs AI agents:

Automation doesn't give you time back on its own. It gives you options. You still have to decide what the time is for.

[Christian | ECHO Consulting @RestrainedDepth]

OpenAI reportó que uno de sus agentes quitó un sistema de monitoreo al copiar código. Cuando le preguntaron, dijo que no lo había quitado.

Por eso a mis agentes no les acepto un "ya quedó". Les pido el recibo: qué cambiaron y cómo lo compruebo.

[Christian | ECHO Consulting @RestrainedDepth]

Microsoft's MAI-DxO, in Science: 83% on hard NEJM cases vs 36% for generalist doctors using references.

As a clinician, what I'd read twice: no specialists in the comparison, text-only, past cases, no live patients.

Strong result. Still a benchmark.

[Christian | ECHO Consulting @RestrainedDepth]

200M+ views over the summer. Watching was step one. Now it’s your turn.

People across 3 continents are already using these tools in 4 languages.

What could you put to work today?

✍️ Turn handwritten work history into a résumé draft. Practice the interview. 🎓 Get organized for college applications and financial aid. ⚽ Explore athletic recruiting and the questions worth asking about NIL. 📄 Make sense of confusing benefits paperwork. 🔍 Find where an AI conversation went off track.

The video shows the actual inputs, interfaces, and results. Pick the tool that solves something for you.

One email opens the showroom. No payment. No password.

Don’t just watch what other people are trying. Try it yourself.

Get free access → https://t.co/DrUM4I6Tlf

[Christian | ECHO Consulting @RestrainedDepth]

Walmart says AI shopping agents will get their own profile in your account: separate order history, approval required by default, optional spending caps.

My rule for my own AI agents: research and prep, sure. Spending waits for a human yes.

[Christian | ECHO Consulting @RestrainedDepth]

Clinical psychology taught me something I use with AI every day: behavior follows the environment.

Vague instructions and a messy folder get you vague, messy work. Same model, clear routine, very different results.