John Repique

I am still calling this TWITTER!

hey world!

🐕this ball is mine #doglife

@elonmusk yep

Try it out!

I didnt know Gautier is still around and still doing his thing…sigh! Can Fashion Still Be Provocative? nytimes.com/2025/10/06/sty… via @NYTimes

Nestled in 20-acres of tranquility, the Cozy Barn Retreat is Shenandoah Valley’s best kept secret! It’s remarkably clean, poshly underrated, and disarmingly relaxing! Ideal spot for honeymooners, digital detoxers, or workaholics on a company-mandated forced vacation .” @Airbnb

we made it to our new home on Tuesday

i’m officially making a come back, sorta

i just baked a Red Velvet 🎂

@seanfenlon @levie I think humanity is not ready for subagents. But at the current exponential pace AI progress is accelerating, the next global divide is centered around AI literacy.

@levie Aw man, social media is not ready for subagents IMHO.

The paradigm of AI subagents is going to be super interesting. There was probably some hope or belief that a universal agent would be able to handle everything you needed in a workflow by stuffing all the relevant context into the context window. But even with larger context windows, this isn’t a panacea, and we see context rot which leads to worse results for long running tasks. So just as we divide labor up in an organization to maximize effectiveness, the same makes sense for agents. The kind of instructions you give a lawyer to review contracts is entirely different from the operations team that processes the contract for a deal. The tools that a security engineer will leverage for testing code are different from the agent that’s creating the code. And so on. Thus, the specific instructions, domain understanding, tool use, and context engineering is necessarily different depending on the unique roles or tasks that you need an agent to do. There are still a ton of questions, like what is the right atomic unit of a subagent - does it get as specialized as jobs are today, or do they own general parts of a workflow (like an agent per section of a codebase)? What’s the best way to have shared state or memory across these agents? How do we ensure “lead” agents can appropriately hand off work to the right subagent to be most effective? Ultimately, we’re in the earliest phases of this paradigm but it’s clear that this will be a major unlock for AI agent effectiveness.

@MmisterNobody @grok context please

There is a darkness in this world, an evil so vast and twisted that the human mind cannot begin to comprehend its depths

In 6 years

Breathing can be intoxicatingly satisfying. #simpleliving #doglife

another glorious day! #HappyEaster

Homemade y’all

When your job gives you purpose and meaning! #TuesdayMotivaton

In mentally healthy work environments, employees can expect to thrive-and-flourish. No need to survive-or-perish! #selfcare #worktolive #notlivetowork

@LizSzabo @NBCNews Malignant catatonia is often misdiagnosed and can be fatal if not treated. ECT is first-line tx. But when patients present in the ED, often it's not immediately recognized and patient 's admitted to another service and not psychiatry.

When talking to people who’ve been misdiagnosed, “one of the things we hear over and over is, ‘The doctor didn’t listen to me.’” nbcnews.com/health/health-… via @nbcnews

@GregoryMcKeown Walking backwards 🤷‍♂️

If you could pick up a new skill in an instant, what would it be? I'll go first: remembering names.

@llmunro The Death of Expertise by @RadioFreeTom amazon.com/Death-Expertis…

@AdamMGrant MLK's dream was perceived divisive in his day because he was an "Original adamgrant.net/book/originals/

It’s often overlooked that MLK’s dream was divisive in his day. Resistance is not a signal of failure. It’s often a clue that you’re ahead of your time. Visions for change elicit opposition from defenders of the status quo. Rocking the boat is a precursor to progress.