Dream job: At @ARPA_H fixing this system.
Can be done. Need the compute.
Small office. The data. Ham sandwiches. Fountain diet coke. I have this whole thing fixed in 2 years - max.
Back in April - I was following the electrophysiology community's response to the CHAMPION-AF and CLOSURE-AF debates.
And then I started doing some math.
A lot of it.
@deredleritt3r@originalmagneto The best way to do this is A/B testing.
Prosecution gets Fable
Defense gets 5.6.
Settle the efficacy of legal research in the courtroom - with real consequences.
Fable may take your job, but 5.6 will never let them take your freedom.
Competitor model releases feel like the iPhone 4 release. But like every week. And I am tired.
This changes everything. Again. And again. And again. And again. And again. And again. Again. We're done. No wait. Again. Ok srsly RSI now. Again. Again. Say it for me one more time. Again.
#solveeverything
Iterate.
Over and over again.
Wake up in the middle of the night with a thought.
Build it.
Wake up later.
Tear it down.
Start over.
When the boots are strapped in ways you simply cannot imagine.
Meta: Part of the reason I enjoy agent orchestration is actually because of the problem. Creativity actually helps contribute to solving the problem itself
The problem that I am working to solve right now is one that has cost me more sleepless nights than I care to think about.
You cannot see it.
It is a problem for hundreds of millions of people (and the people who love them).
The problem presents itself differently and similarly - at the same time.
People are losing their lives because of this problem. Not just in the sense of life or death. But in the textural sense of life.
It is not a profitable problem to solve. Yet, it is necessary to solve.
This is 10 years in the making.
OneRhythm.
#solveeverything
@ryanbrewer Books were cool.
The internet was dope.
Bandwidth increased.
The internet became doper.
...and then
LLMs - the dopest way to learn anything.
At present.
Const consequence analysis for a medical device in cardiac care (some pretty alarming data)
Im going to publish that research paper. The industry is likely not going to be a fan of me. I don't mind. I work for patients.
Then, I am following up that paper with an article that narrates the research paper - and what is at stake.
Then - my number one priority is to get a peer support platform that I have been building into production. It's about 90%. I care about this project more than anything I have ever worked on in my entire life. Personal for me.
Then, once I have access to the data, I have some deep learning work that I'd like to begin. The work is aimed to identify psychological distress markers for cardiac arrhythmia patients across clinical and consumer wearable data.
Once all of that is done - then I will rest.
Until then. Full send.
Indeed - I use AI mostly for research anymore. Codex has been clutch for me in generating scripts to analyze large data sets.
I am having the time of my life with it tbh.
Things that would have taken me weeks in the past are done for me now in hours.
It is kind of mind-blowing.