Hacker News Reader

Comment by asah | Hacker News Reader

People have struggled with this for time eternal, and from Google 20% time and other "skunkworks" projects, some of the greatest discoveries and inventions were made off the books.

One of my favorite (1993):

Wiles, with his from-childhood fascination with Fermat's Last Theorem, decided to undertake the challenge of proving the conjecture, at least to the extent needed for Frey's curve.[18]: 226 He dedicated all of his research time to this problem for over six years in near-total secrecy, covering up his efforts by releasing prior work in small segments as separate papers and confiding only in his wife.[18]: 229–230

https://en.wikipedia.org/wiki/Andrew_Wiles

View on HN

I have heard a similar story about Richard Feynman. If I recall correctly, Feynman claimed to have only applied for grants on work that he already completed, that way he knows he can fulfill what he set out. I am unable to find the source for this but I presume it was in his autobiography.

This is a common trick for theorists. You need to have essentially completed the work in order to write a compelling grant proposal.

I did this when I was a researcher even though I was not a theorist (I was working in mobile robots). I always worked a year ahead, and write a proposal for what I had done the year before. I acquired a reputation as someone who always delivered. Then one year I had a proposal rejected on the grounds that what I was proposing was impossible to achieve within the budget and time frame I was proposing. That's when I quit.

They wanted you to ask for more time and money? What kind of grift were they actually running?

People evaluate the proposals expecting them to be estimations. And when somebody estimate some work taking too little time or money, people often don't trust that estimate.

This is all completely normal and reasonable. Except for the part where you are expected to estimate results of scientific research.

Yeah too good to be true things are red flag for fraud or pipe dream

I hope you quit, took 6 months off, then came back with the previously finished work and told them, “You were right! It only took six months and zero grant money; I did this while traveling in Japan.” That might have blown their little bureaucrat minds.

The problem is that the reviews were anonymous (which was supposed to make the process less political). My direct managers knew what I was doing.

it's not uncommon for labs to ahve a grad student working on a skunkworks project funded by a previous, unrelated grant. They generate initial data, which is shown in the grant application. But often the grad student is still working in parallel with the extremely long grant cycle to get more publishable results and can sometimes even get to a full paper before being funded.

I think of it like speculative execution and pipelining.

That's the sad reality of most grants today

I created and sold a software system to a company and joined them. I stayed there for 6 years continuing to work on it and building a big team. During this time, I quietly spun off a small rogue effort to build a newer client app for the system in Rust. This quickly became our most popular product.

The company is now winding down the project overall, but decided to keep and adapt this Rust client to its own backends. It is the only part which survived and it was the only thing I wasn’t given orders to build.

Harder questions need to be asked regarding how metrics tie to actual objectives. Just because I have a chart trending in the right direction is meaningless. “Number of papers published” is not a good metric.

We need more trust and fewer metrics / games. The thing I really don’t want to pay for (as a tax payer) is the bureaucracy in the middle of all of this.

Might be a good idea to bring back the concept of the gentleman scientist, who are generally independently wealthy. Charles Darwin is one such scientist.

Great, I think it'd be so much better if a qualification for scientist was "already rich".

maybe, but it's a hell of a lot easier to get rich when you don't spend your 20s on a phd working for peanuts

As a researcher who is seeking alternatives to the pressures of both academic and industrial research, one of the things I’m thinking about for my long-term career plan is to relocate to an area with a lower cost of living (bye Silicon Valley) and make an arrangement where I work part time for money (possibilities include being a contractor, doing freelance or consulting work, or starting a business that doesn’t require more than 20 hours per week to operate) and spend the other 20 hours a week on research. I’d love to be able to do research based on my own interests without having to worry about business or career concerns. I’ve come to realize that when one is paid to do research, the researcher is suddenly under pressure by the person or institution paying the researcher. “He who pays the piper calls the tune,” after all. But if I make a living through non-research tasks, then I’ll have research freedom.

The tough part is coming up with a business idea that would allow this type of lifestyle; that’s where I’m currently at. If I could make $130,000 per year, heck, even $65,000, and live in an inexpensive part of America then I’d be fine, but the trick is figuring out how to make that much money in only 20 hours instead of 40.

Why not live somewhere even cheaper? Do your research from a nice hut on the beach in Thailand for example.

What do you do research in?

I work in machine learning, and I also have interests in programming languages, operating systems, and databases.

Working in research, I have observed that one can't do everything alone, so many things effectively need a team because specializations matter and some tasks simply need a decent quantity of mundane effort. So me living cheaply wouldn't be sufficient because I still need funding to pay all the other people involved.

It seems that we are getting short of people interested into paying with their own money for the privilege of solving other's people problems for free. Is very strange.

It's a hard problem because there are no good answers.

> We need more trust and fewer metrics / games.

Except, at scale when dealing with a lot of people, there are plenty of people who are willing to abuse that trust.

I made this mistake early on in my career as a manager. I gave an important piece of work to one of the very senior engineers who I managed. Over the course of a month, he repeatedly said "trust me" regarding the status of his project. When 6 weeks went by and he finally showed it, it was in such a state that it was essentially unusable. We ended up throwing the whole thing out. Furthermore, he left for another job about two months after that, leaving his work in a shitty state.

I made plenty of mistakes there, but I also learned a valuable lesson, "trust but verify". Until someone shows me concrete evidence of what they've done, I really have nothing to go by.

To be clear, I'd love to be able to "have more trust and fewer metrics". But at the end of the day, it's really hard to give people the freedom to, say, hole up for months at a time while they say "trust me".

Managers need to have the capacity to understand the work being done. Get into (I assume) code, and see what is happening. If you can’t do this, it is impossible to be an effective leader. You can be an effective people manager without this capability but you cannot be trusted to ensure the work will get done.

Totally agree. To be clear, I am technical, and I definitely was able to evaluate the work being done once it was delivered. My mistake (among many) was letting the work go too long without any visible progress.

That's why I get annoyed when people just lament about the problems with things like OKRs and KPIs, while totally disregarding the problems these things were designed to solve and pretending that if managers just had a more "trusting" mindset that a million Higgs and Wiles would be free to blossom.

I don't know what to think about a 'senior' engineer that doesn't reflexively keep his manager in the loop.

I don’t understand the “once it was delivered” aspect. Why not ask people to commit whatever they have and then you can see if it is going in the right direction (or not).

If you can’t do this yourself, at some point you need to be able to trust (yes trust) someone who can. You could not formulate OKRs or KPIs that would have solved this problem.

That's exactly the "verify" your parent was talking about. Unless you happen to have a situation where your project includes regular submission of work artifacts -- which describes a lot of modern software, sure, but is far from universal -- monitoring the work regularly is at odds with the "trust" you mentioned above. I've often seen managers who attempt to monitor like this accused of micromanaging.

You always have to remember to feed the KPIs, otherwise the management boogie men will take away your health care.

This is why you FIRE, so management loses their control authority and you can actually focus on solving problems without their interference.

Healthcare coverage is tied to reaching KPI's in he US?

I find myself doing this for tech debt items far too often.

Reply on news.ycombinator.com