[neilv]

The article names its archetypes not entirely fairly, based on my experience at three MIT startups. Two were novel-tech academic spinoffs, and the third was alumni applying/integrating COTS tech driven by customer problem domain. Of those three, two were business-thinking much like the article's "Stanford" archetype. The other had, like the article's "MIT" archetype, a period to develop tech from the lab, and a challenging changing of gears to be cutting-edge product-driven.

Also, a fourth MIT startup, which I recently almost co-founded, even before I joined the nascent team, the inventor-CEO had already done impressive customer-oriented legwork, and found Bay Area advisors, more like "Stanford" in the article.

> The team has 9 PhDs and just hired an MBA to start finding customers.

This is a problem, if none of the PhDs happen to have non-academic strong experience in product, nor in industry team engineering.

An academic environment will tend to make people think they know more than they do, about things academia doesn't know.

Also, academic degree and career paths in some ways reward the opposite of how I think people in a startup, or other effective company, should be thinking. (Unless the startup is more the VC growth investment scheme kind, which can be mostly about appearances.)

One MBA (even if very experienced) probably can't, by themself, counterbalance all those experience gaps, nor those lessons to unlearn.

> The MIT startup has no sales to customers, but possibly a DARPA grant to develop their technology.

I don't know about the more involved DARPA grant-writing, but SBIRs do seem to be popular seed-ish funding: https://www.sbir.gov/

[sybercecurity]

SBIRs can be a good first set of funding. Some agencies go with $100k for six months, so win a couple of grants and you are doing pretty good. The problem is that some agenices have different criteria, forms, etc. so just because you jumped through the hoops for one funding source, it doesn't automatically translate (government does love their paperwork/processes).

Some programs are specifically designed for commercialization and want to see a product available at the end. There was a big shift to that post 9/11 (sometimes called "little 'r', big 'D'": less research, more development.

[choppaface]

And Google is a pretty good counter-example to “The Stanford startup has developed no new technology”

[Nevermark]

(Google is huge, and does real tech, i.e. advances in deep learning. The following only applies to their search startup phase.)

PageRank was innovative, and had great "market fit" (or "user fit", prior to ad revenue), but wasn't very deep as technology goes.

Google's (original) business success came from leveraging that into a three-way network-effect as the middleman connecting web content users, web content providers, and web advertisers. Even today, PageRank's successor algorithms appear to be aimed more at ad revenue optimization than improving search as a technology.

Likewise. Netflix, Facebook, Expedia, Spotify, Uber, Airbnb, ... use tech as operations, it is not the product. Their primary defenses are not unique technology, but network effects, customer information lock-in, and other market-side moats.

They have great technical people doing great work. But a high proportion of their innovation is aimed at their own operations, customer engagement, etc., not creating and offering new technology.

--

Technology first companies typically end up as part of a supply chain.

nVidia is a good example They have become huge, but are still a parts maker in a lot of ways. They have created a sticky ecosystems, but their primary work and moat remains keeping up a relentless technological cadence.

Another example was Amazon. They started by adapting others' tech to create an online store. But instead of simply optimizing that, they fully developed distributed computing as a service, and spun that out as its own business. Now they are a provider in other companies' supply/resource chains.

[beowulfey]

I don't think these archetypes apply to the mid-90s Stanford/MIT scenes

[jjssmith]

AFAIK, the Markov chain page ranker existed before Google (Jon Kleinberg's papers, e.g.), what Google paper added were some smaller bells and whistles, like using the text description of the hrefs.

[Symmetry]

SBIRs can be useful, but they can also be a trap where a company just keeps going after SBIRs and never takes the technology to market.

[beryilma]

There are entire companies in Boston/Cambridge area whose business model is to get perpetual SBIR grants back-to-back and never produce anything of commercial value. And they pad themselves on the back for being innovative companies. It is almost like there is collusion between these companies and grant-giving agencies.

[pbronez]

Yup - “SIBR Mills” The Federal government has very complex and specific rules for buying things. It’s a big, important market but hard to crack. Thus the most successful players tend to be really good at selling to the government… regardless of whether they’re really good at what the government needs.

The government knows this and thus tries to create programs that streamline the process while retaining sufficient controls to steward tax payer funds effectively. SIBR is one such program. They’re useful, but not immune to exploitation by sales specialists.