Science Degen

Ex-scientist, still curious. Science is cool, data is cooler, but only when it’s not playing hide and seek with the truth. Question everything learn on 𝕏

[Science Degen @ScienceDegen]

@rileybrown 100% accurate. I live in folders that push data to Codex/Claude, that accomplish my tasks. If you can't monetize the app, then building the skills connected via MCP is the next best move? I'm not monetizing squat, but my worklfow is as you describe and I'm not going back!

[Science Degen @ScienceDegen]

The strong version of that claim is too confident: scientists have not universally “confirmed” that unmedicated adult ADHD causes anxiety as a deliberate compensatory mechanism.

The more accurate extrapolation is:

Some adults with untreated or undertreated ADHD appear to learn to use anxiety, urgency, shame, deadlines, or fear of consequences as an external arousal system. That stress response can temporarily raise physiological activation enough to initiate action. It can work, but chronic reliance on threat-based activation may worsen anxiety, sleep, avoidance, burnout, and executive dysfunction.

ADHD is not simply “low dopamine”; it is often a problem of regulating arousal, salience, motivation, and executive control. A systematic review found autonomic nervous system differences in ADHD, often leaning toward hypo-arousal, though findings were mixed; stimulants and rewards can increase arousal in some studies. Separately, prefrontal cortex function follows an “inverted U”: too little dopamine/norepinephrine can impair focus, but too much stress chemistry also impairs it.

So the mechanism is less:

“Worry gives the ADHD brain the dopamine it lacks.”

And more:

“Threat makes the task salient enough that the nervous system mobilizes.”

Anxiety can create urgency. Urgency recruits norepinephrine, dopamine, cortisol, heart rate, vigilance, and threat attention. That may help someone start the thing they could not start when it was merely “important.” But it also teaches the brain: I move only when danger is high enough.

The cost is real, but “running on cortisol as fuel” is metaphor. Cortisol is not fuel; it is part of the stress-response system. Used occasionally, it helps mobilize. Used constantly, it can mean the person is living in a loop of procrastination, panic, overperformance, crash, guilt, and repeat.

A sharper version of the statement would be:

Adults with untreated ADHD may develop anxiety-driven productivity as a learned compensation. Because ordinary importance does not reliably generate enough activation, the brain may come to rely on urgency, fear, or impending consequence to start. This can be effective in the short term, but over years it can turn motivation into threat management and leave the body chronically stressed.

Sources: adult ADHD commonly co-occurs with anxiety and may be under-treated (BMC Psychiatry https://t.co/PnOp8ME7Pj); ADHD shows evidence of arousal/autonomic dysregulation, though not uniformly (📷Bellato et al.https://t.co/PnOp8ME7Pj); stress chemistry has an inverted-U relationship with prefrontal function (📷Arnsten review https://t.co/RNt9ujqW7s).

[Science Degen @ScienceDegen]

@cybertruck @Tesla I want you but I would need to exchange the 2021 X, which my wife drives. Can you come over, kidnap and exchange? The X, not the wife....

[Science Degen @ScienceDegen]

@WhitePapersPol Those of us that pay taxes are done with this.

[Science Degen @ScienceDegen]

@GovSherrillNJ hmmm....

[Science Degen @ScienceDegen]

@cpsj As I continue to discuss with my wife...the only solution is to leave NJ.

[Science Degen @ScienceDegen]

@svpino Back in the day, this is exactly how I made novel discoveries in the lab. AI is allowing me to break through and learn at a level I never thought possible.

Stay curious—or else. That's my hack. Always curious...

[Science Degen @ScienceDegen]

@jalehr I might know a mammal...

[Science Degen @ScienceDegen]

@MatthewBerman Yep, no more oauth...I was just beginning to warm up to them. I need my own GPU's...

[Science Degen @ScienceDegen]

This is right but missing a step. Before you "build an AI tool that does it 10x faster" you need to know which endpoints exist and how to chain them. I run 60+ automated skills — SEO audits, competitive intel, PubMed research, FDA monitoring — each one started as a client deliverable, got scripted, then became a tool. The pattern: manual work → script → skill → product.

[Science Degen @ScienceDegen]

@veggie_eric I tested it with the Bible. Next FDA docs.

[Science Degen @ScienceDegen]

@ryancarson @lennysan @openclaw antfarm is awesome!

[Science Degen @ScienceDegen]

Workflows. For example: Transcribe video from X → download to MP3 → run Whisper. Research. Another set of logical steps that OpenClaw follows and produces the output I desire.

What is your task? How do you get the data? What do you do with the data? If you can explain that process, it becomes code, and your mundane tasks get automated. That is OpenClaw in a nutshell (oyster shell).

[Science Degen @ScienceDegen]

@DavidOndrej1 Best quote: "Learning how to verbalize your thoughts is going to be one of the most valuable skills in the coming years"

[Science Degen @ScienceDegen]

OpenClaw is awesome when connected to my data. I teach it scenarios and write scripts to help me achieve better outcomes. It has become my assistant. I fear I'm replaceable by code... Cool, can't wait to learn more.

Stay curious, or else.

[Science Degen @ScienceDegen]

@DorianDevelops @openclaw I'm testing on old MacAir and old Ubuntu from 2017. You don't need power if you are making cloud API calls.

You want to buy a computer? Buy the Studio or Spark and run local models. That upfront cost might save long term on buying Opus tokens?

[Science Degen @ScienceDegen]

@EXM7777 I would use them to build awesome educational assistants that allow young, unbiased minds to learn and build the future.

P.S. I would prefer 1 Mac Studio or 1 DGX Spark over 10 Mac Minis. The ability to run huge open-source local models is key.

[Science Degen @ScienceDegen]

@rileybrown @designertom I am loading in 'skills' in the form of python code snippet examples, in the form of API endpoints mapped to scenarios that perform my work.

I am teaching it how I do the mundane and complex tasks. I am cloning my knowledge into code. I am replacing me and creating Me+

[Science Degen @ScienceDegen]

@ryancarson @openclaw I think the only reason to chase a shiny new box, is if it has the power to run the best open source models. ATM $4K-$8K in hardware can get you there?

I asked my wife and she said no :(

The cost of using OpenClaw via API to Opus is nuts.

[Science Degen @ScienceDegen]

There are 2 types.

My known knowns: my verified mundane flows of logic and patterns—my truths, if you will. More than a soul, these are my https://t.co/OzCnXxQPZv files. The API endpoints mapped to potential solutions.

The second type is client-specific. Tone, voice, etc., are more of a soul input. However, I like to load the scrapes and metadata behind the scenes manually. The more I can curate and QC, the better my OCD feels.

Am I over-engineering things the bot could figure out? Perhaps. Learning as I go, but the knowledge bases are a big part of how I'm setting up my current "work" clone.