How would I improve American scientific output for $100B? A thread. 1/n

First, we already have a lot of processes that seem to work incredibly well, and processes really matter! Institutional knowledge matters! So whatever you do, _don't break existing things_. 2/n

Talent is a massive problem, and we are sending, at best, mixed messages to people internationally. I'd take $10B and buy off various anti-skilled-immigration constituencies. 3/n

I'd then split say $50B of the remaining $90B across existing agencies (maybe proportional to current size) but stretched out over 15 years. The NIH funding bolus in the 90s caused lots of structural problems! 4/n

I think we should also dump considerably more money into funding _individuals_ over long (say, 5 or 10-year) time periods, ala HHMI investigators. Fewer grants, more independence, more opportunity for curiosity. Call this $5B 5/n

I would ramp up early-stage-independence awards, like K99 on steroids. It seems that a lot of scientific progress occurs at the margin by lots of people trying lots of random things. Let's have more independent attempts, earlier! (eve if it means fewer giant-collaborations!) 6/n

Maybe start those small, a few dozens per area per year. so maybe $200m/yr or $2B over 10 yr? Oh, also buy off whoever you need to to make everything open access going forward. So tired of having this fight. Say that's $100m/yr so $1B total. 7/n

I really think we should invest in instrumentation/automation and replacing labor with capital. Right now a lot of funding agencies make it hard to pay for anything OTHER than labor. This doesn't scale and creates pipeline problems! Janelia has the right attitude here. 8/n

It seems like the BRAIN initiative really had the desired catalytic effect on instrumentation development, although that might have been a right-place-right-time phenomena. Still, a lot can be learned here. 9/n

Academic compute infrastructure is a massive nightmare and is dragging _everyone_ down. Access to compute is abysmal even at good places. I'd give Bezos $1B/yr for AWS Science Cloud. Over 10 yr that's another $10B. 10/n

Are we getting close yet? This doesn't even address some of the structural changes that are hard to address with pure money. But my final thoughts are: 11/n

1. Don't throw baby out with bathwater, scale what works.
2. Talent talent talent.
3. Create systems that enable long term commitment to science by individuals
4. less paperwork, more science done by scientists
5. More infrastructure! /Fin!