In order to do that, lets walk through the pieces of the niche one by one and examine how a private ARPA could make sure that it’s structurally set up to cover each piece.
This will require explicitly and uncomfortably pushing back against the (accurate) common startup wisdom to “focus on a niche.” The more you follow startup wisdom, the more you’ll be bound by §Startup Constraints. At the same time any technology does need to find an initial niche eventually and a program should have some precise hypotheses about an eventual application. Counterintuitively, this tension between considering how a technology will be used and potential local-optimum traps<Tensions are the core theme of the PARPA writeup> may be one of the barometers for a healthy organization, ensuring that it doesn’t lean too far one way or another. Maintaining this unstable balance will require a lot of work to frame goals correctly. While cliche, the development of the transistor is a good example - it had a precise goal “find a solid-state replacement for the vacuum tube that we think will be useful in telephone repeaters and switches” but didn’t constrain the work to only “make more efficient repeaters and switches.”
The structure of an ARPA-riff can enable targeted dicking around through low-stakes seedling programs, having programs exist under a single organizational umbrella, and having some number of ‘free radicals.’
An ARPA-riff that copies DARPA’s use of low-stakes seedling programs<DARPA PMs use seedling programs to ‘acid test’ the riskiest pieces of a program idea> can explicitly go into them with the attitude of “let’s see what will happen.” It’s important to acknowledge that of course, everyone would rather the experiments point to a promising program than be a nothingburger, and the way to prevent that preference from morphing into pressure is to explicitly make sure that negative results don’t have lasting effects. Specifically, this means that negative results can’t affect a PM’s ability to run other experiments or affect a performer’s relationship with the organization as a whole.^1
An organizational umbrella over programs can prevent serendipitous results from evaporating which can make targeted dicking around beneficial over the long term.^2 If each program was a separate organization, pursuing any serendipitous discovery that didn’t help it move towards its goals would be a distraction. Someone might keep track of them but there would be relatively little incentive to do so.^3 An umbrella structure would enable serendipitous potential side-quests discovered during the main body of a program to be fed back to the beginning of the “pipeline” and possibly become its own program.
An ARPA-riff could also enable targeted dicking around by explicitly having people whose job is to do targeted dicking around. These “free radicals” could be competent individuals who aren’t attached to any specific project. They could help PMs design programs, do experiments, or research areas that nobody in the organization has even thought about. You could imagine this role as having the flavor of a “pre-PM X-in-residence” where it is like an internship in that the person isn’t doing anything mission-critical and could potentially lead to them becoming a PM without the expectation of that outcome. It would be unlike an internship because it would involve much more independence and the people in the role would be expected to already be skilled so there would be less of a teaching component. This sort of role is anathema to efficiency and could easily be a waste of resources, but rhymes with how Bell Labs had researchers who were ‘between projects’ and would help out other researchers or poke at possibilities.<Bell Labs enabled free radicals>.^4
The first way a private ARPA could fill this niche is through interactions between program managers. Program managers should be drawn from a wide range of technical experience. This diversity combined with setting up cultural expectations are set up correctly (riffing off of Bell Labs’ open-door policy^5 for example) could go a long way towards fulfilling this part of the niche.
It would seem like programs that initially revolve around externalized research might present a challenge to creating high-collaboration environments. However, DARPA PMs seem to do a good (emulatable) job addressing these challenges by holding regular closed-door meetings (so people can talk about real problems instead of just showing off good results) and making sure different performers know each other.<A large part of a DARPA program manager’s job is focused network building> ^6 The shift from externalized research to internalized over the course of a Exploratory Program Organization will make the program further look like parts of a corporate lab.
Realistically, a private ARPA won’t be able to fulfill this role as well as a golden-age corporate lab, but it will fill this part of the niche better than academia or startups do today.
Seedling programs, encouraging serendipity, and free radicals (see ::put link to #2::) as well as running multiple programs with different levels of maturity<A private DARPA riff both can and must do multiple experiments at once> could go a long way towards enabling smooth transitions from early TRLs to middle TRLs. Admittedly, smoothing the transition from middle TRLs to later TRLs is tricky to structurally build into a private ARPA. One potential smoothing mechanism is how Exploratory Program Organizations can become independent organizations whose purpose is to get the technology that they developed out into the world, either by selling it as a startup or helping people adopt it as a non-profit. Of course, many technologies are more effective if they're absorbed by existing organizations — especially things like new processes.^7 Building a consistent system to encourage existing organizations to integrate new technology is incredibly hard — even DARPA has a mixed record on transitioning. Despite that difficulty (and perhaps because of it) it’s something that an ARPA-riff needs to strive for from day one.<An ARPA-Riff could establish sales channels for frontier tech>::This feels like an awkward link::
Similar to the transitions to existing organizations mentioned in #4 above, this is a challenge for an ARPA-riff compared to a corporate lab. There are several experiments that the organization can and should try in order to establish default customers<An ARPA-Riff could establish sales channels for frontier tech> but they are all contingent on relationships, reputation, and specific circumstances. Realizing that positioning technology to fail is a potential failure point<Reasons why an arpa riff would succeed or fail> will hopefully make it into a self-defeating prophecy but at the same time, too much focus on adoptability can push programs towards boring safe decisions.
Not being officially attached to a parent company makes an ARPA-riff by default worse at this than a corporate lab so it’s especially important to explicitly set up these feedback loops wherever we can. However, feedback loops are a piece of the golden-age corporate niche where an umbrella organization over a number of programs is more useful than a set of stand-alone programs^8. Consistent cycling of PMs with a diverse set of industry experiences provides one feedback loop — it suggests that ‘industry experience’^9 should be something to look for in PMs.
There are a number of other (hand-wavy, but worth noting) ways to set up feedback loops between an ARPA-riff and real problems. Like Ink and Switch, the organization could do consulting work to help external organizations implement new tools/processes. In addition o feedback this sort of consulting would both help address #5 above and provide an alternative revenue source. An ARPA-riff could set up a consortia similar to the Media Lab, which would give the organization a default reason to talk to other organizations.
Especially early in their lives, a private ARPA’s programs will still lean heavily on academic work, so the organization won’t address this part of the niche as thoroughly as corporate labs or Focused Research Organizations. However, if the idea of ‘gradual internalization’ works, private ARPA programs<Exploratory Program Organization> could become an alternative to academia for some people, some of the time.
This is a place where organizational discipline is important. The temptation to go full “we’re going to fix research!” is strong but I suspect that attempting to provide a first-class alternative to academia would undermine many of the things that make the ARPA Model work in the first place. Instead, the organization should support other alternatives to academia by not biasing towards researchers with academic affiliations. While a single organization is probably unable to offer a first-class alternative to academia, one could dream very big and imagine that an alternative could emerge from an ecosystem of ARPA-riffs and other new types of research organizations. This is yet another reason why PARPA should create a replicable institutional model.
One might expect that it would be hard to support long-term projects in the ARPA model because PMs with relatively short tours of duty is a core feature. However, DARPA has a strong precedent of programs outlasting the tenure of PMs who started them. There’s no reason a private ARPA couldn’t do the same thing. Arguably, PM-PM handoffs can maintain continuity better than many academic labs where most of the work is done by grad students with ~5-6 year tenures with very little incentive for smooth handoffs^10 or startups, where employees are notoriously mercenary and it can be surprising if someone stays more than three years.
However, it’s also important to consider constraints that the organization will face early in its life that will hopefully relax over time. While an ARPA-riff is structurally set up to tackle long-term projects, it’s probably not a good idea for the first projects to take 6+ years to show results. The reality of the situation is that a new organization will have no track record and so people will (rationally) have less patience with it. Organizational reputation matters. If funders, potential employees, and collaborators start to see the organization as unable to get things done^11 it can become a self-fulfilling prophecy.
There’s not much to say here besides this is the tightrope that an ARPA-riff should try to walk: things that are too research-y for a startup but not paper-focused enough for academia.
^1: To paint the counterfactual of explicitly making sure experiments have no negative consequences: One could imagine that after a failed seedling experiment a PM could implicitly or explicitly need to ‘redeem’ themselves by going above and beyond. Similarly, the organizational attitude towards a performer who ran a failed experiment might implicitly shift towards “well they’re good people but a bit incompetent.”
^2: As opposed to something like flossing where people are like “yeah I really should be doing that … later.”
^3: This is an example of the more general principle that Optimizing individual components of a system is often at odds with the system itself.
^4: These are the ‘fellows’ in Phase Two of Pathways towards PARPA.
They were not to work with their doors closed. They were not to refuse help to a colleague, regardless of his rank or department, when it might be necessary. From The Idea Factory.
^6:Focused network building is a valuable early step for riffing on ARPA
^7:<sidenote> See Fundamental Manufacturing Process Innovation Changes the World to expand on this point.
^8: <sidenote> See Adam Marblestone and Sam Rodriques’s Focused Research Organization proposal.
^9:As suitcase handle-y as that term is.
^10:”I’m busy writing my thesis. Here’s my lab computer - it has everything on it. Good luck!”
^11: People absolutely run out of patience even if you set expectations up front. https://twitter.com/Ben_Reinhardt/status/1350507766381744128