Category: Ideas

I, For One, Welcome our New Admissions Overlords

*To paraphrase Kent Brockman

Once upon a time, there was a yellow brick road that led to college. You would submit your SAT scores, your GPA, your activities, you would write your essay – and you would submit all of this on paper! And all was good and just in the land, and all of the right students gained admission.

Utter nonsense of course! The college admissions process was a mess then, it’s more of a mess now, and it’s about to get hit by one neutron bomb everyone’s talking about (SCOTUS case which likely ends affirmative action), and another that may be even bigger.

But first, a blast from the past – I was among the last classes of students to apply on paper (Dec 1994) – perhaps one good thing about that era was that students applied to fewer colleges, since they couldn’t shotgun their application to 20 schools via the Common App or the internet. I applied to 6 schools, and had the good fortune to get into all but one. My safety school at the time even had programmatic admissions – if you had a GPA above X and an SAT above Y, you were essentially guaranteed admission, not just to the university but to the honors program!

Admissions have gotten harder since then, although the numbers are a bit of a lie, as elite schools try to lower their acceptance rates as part of the college rankings game – they use the ease of the internet to lure unsuspecting students to submit applications that have no chance of success. Admissions have also become less structured, as more and more schools have eliminated or de-emphasized testing requirements – with the occasional retrench, as my alma mater reinstated the SAT as a requirement (going to great lengths to explain that it IS actually correlated to success at an engineering school). But all of these changes pale in comparison to 2023…

College admissions will be impacted by the end of affirmative action. But they will also be deeply impacted by the rise of generative AI! I’m willing to bet that numerous high school students used ChatGPT to help write their essays this past December. And even with the new paywall, ChatGPT and its competitors are far cheaper than the pricey consultants that wealthy families use for essay ghostwriting (let’s just acknowledge that this happens). While colleges will attempt to deploy tools to stop the practice, students aren’t that dumb – they’ll add their own touch to the essays, making it hard to tell where the robot dropped the pen and where the student picked it up. So what happens to college admissions in an environment where affirmative action is dead, standardized testing is diminished, and essays are written by ML bots?

In the spirit of my days at HiddenLevers, here are a few potential scenario outcomes:

Back to Basics: Schools (in collaboration with SAT/ACT) will reemphasize controlled measures like standardized tests, GPA, class rank, and similar, since they can’t trust much else. This will dismay some and delight others, but it’s easy ground to tread since this was the norm not so long ago.

Human Interviews: Zoom eliminates a lot of the costs of the traditional college interview – but instead of using it as a “positive” tool, schools may begin to use it as employers do – as a primary filter mechanism. This approach will lead to a wide variance in outcomes just as it does with corporations (some are good at using interview-based recruitment to acquire talent, and some simply suck at it).

Welcome Robot Overlords? Here’s a guess for a post-affirmative action AI-enhanced world: admissions decisions will themselves will be handed over to machine learning. By placing a black-box trained algorithm as an intermediary between themselves and admissions decisions, colleges will achieve several goals:

1) Algorithms will likely achieve a higher fit toward whatever class composition the administration wants than human admissions officers. Simply feed in past classes or “idealized” classes and let the algorithm build such a class from the applicants. Colleges will take this approach because…

2) This decreases perceptions of bias by offloading human biases into the algorithm’s training (which leaves a lot more plausible deniability, regardless of the goals!)

3) The overall cost of running admissions, even with human oversight, will be substantially lower. And since we all know elite universities are just hedge funds with educational arms anyway…!

It’s an unsettled time for colleges and prospective students, and this post hasn’t even touched on issues like falling birth rates or the rise of alternative career pathways! But post-secondary institutions are going to have to wrestle with the impacts of ML just like the rest of us. In some situations there are clear right or wrong answers, but that’s not the case here. Would it hurt to put a robot in charge?

Posted on by praveenghanta 0

Embodied Time

The New Year makes many of us think about time. I got to thinking today about embodied time, human time encapsulated into physical form. What is embodied time?

We call it money. We use money to buy things that we don’t have the time (or ability) to produce ourselves. Given infinite time, even I could learn how to grow cotton, harvest it, gin it, and weave it into shorts – the shorts I needed for a run today because I forgot to pack them!

If I had endless time, perhaps I too would follow the Gandhian tradition of making my own clothes. But I don’t, and I chose to use embodied time, aka money, to buy those shorts from someone else – in this case a large corporation. At the end of the day that corporation applied human time and effort to the creation of those shorts.

We apply modern technology to amplify human productivity, and the linkage between money, time, and the economy can be seen in the concept of Productivity. In economics it’s measured as $ of GDP per hour of labor – it quite literally expresses to us that money is embodied time! Perhaps AI and robots will free us from this linkage, but recent advances show us just how far we still have to go.

The economy today is constrained by a lack of time. Not enough humans are able or interested in investing their time in production, and this is the chief obstacle in bringing inflation to heel [1]. If the economy is constrained by time, that’s because we as humans feel constrained ourselves. Retirees cannot and do not want to spend their remaining years working, and core population growth has gone negative across the developed world – there just aren’t enough new kids on the block.

What’s to be done? We need to find more human time, more work hours. It turns out that capacity is right under our noses! An increasing percentage of working-age folks aren’t working, but it’s not at all uniform. A substantial percentage of full-time workers want MORE work – these are the workaholics that want 60 hours a week long term. Hey that’s me too, and I recognize them because I’m part of that crew. I’ve averaged 60 hours a week for most of my career, and many startup founders and employees are quite familiar with these sorts of hours. And medical residents would scoff at how easy a schedule that is!

When I started fraction, it was based on a core hypothesis: that 10% of the US professional work force both WANTS and is able to work 60 hours per week. This hypothesis has proven out – about 40% of developers, designers, and product managers that we’ve spoken to (a thousand and counting) are interested in taking on extra work, and roughly one-fourth of those have the capacity and ability to do so. 10% of the 80M US workers able to work remotely is 8M part-time workers, or 4M FTEs. When I validated this, I felt like a gold miner sitting atop a massive gold seam – in an economy desperate for workers’ time, we’ve got labor time enough to close the entire US labor shortage!

I’d like to take those workers’ time and embody it, turning it into money (and economic benefit) for US workers and companies alike. A lot of folks I know wonder why, after a successful saas software exit, why on Earth would I get into a services-oriented business? My answer: I feel I’m on a quest to supply the resource the country needs most today – skilled labor. More knowledge work getting done means more solutions to every problem today, from climate change to health care to Americans’ top current priority, inflation. I’ll plow all of 2023 into that, and hopefully in the process help a lot more people embody a lot more time.


[1] I’m generally optimistic on inflation given the trends seen in the energy, food, home, and goods sectors – the Fed’s moves are working. But they seem determined to keep the heat on to corral wage inflation as well, I wouldn’t be surprised if we see some negative prints in the coming year.

Posted on by praveenghanta 0

Abuses of US Non-Profit Status

The holiday season is typically when US charities collect the majority of their donations. Let me start by noting that the US non-profit sector is vibrant, generating positive outcomes across many aspects of society. But there’s an unfortunate dark side, and a degree of abuse of the system that is appalling. When organizations with CEOs making $10M+ per year spend little on charitable services and pretend to be non-profits, that’s appalling. When organizations that are blatantly political in nature raise limitless funds and pay no taxes, that’s appalling. When some of the largest hedge funds in the land pretend to be “universities” while raking in billions, that’s appalling – and the list goes on.

I have given to charities for many years, but with the start of our family foundation, I’m now formally involved in the sector, making choices about grantees. The process of grant-making has enabled me to refine my thinking on the kinds of impacts we’d like to see, and on how to measure and quantify our impact. But seeing a range of pitches and organizations also helps me to see and understand what I don’t support – this rant could run a mile, but here are the quick hits, the sorts of organizations that I believe fundamentally abuse the tax code:

Non-Profit US Hospitals: The majority of US hospitals are organized as non-profits, and yet non-profit hospitals spend only 2.3% on charity care, which is actually less than for-profit hospitals! Non-profit hospital chain CEOs have compensation similar to Fortune 500 CEOs, and don’t do much charity work, so why don’t we end the sham and strip their status? Senator Chuck Grassley has been a lonely voice of reason on this issue for years. Ask yourself – why is it that doctors, pharmacies, medical labs, pharma companies all pay taxes – so why do hospitals get a special free pass?

And what’s more evil than pretending to be a non-profit while bankrupting patients by the thousands? Doing it while using sick kids as a way to tug at heart strings. Unfortunately most children’s hospitals also provide very little charitable care, and also pursue patients into bankruptcy just like any for-profit company.

501 (c) (4) organizations: These organizations were original meant to be “social welfare” organizations, but have now devolved into another form of super-pac which lobby and influence politics via donated funds! But why aren’t they taxed on “profits” like any other corporation? Both parties have used and abused these structures to the max now – pray for us poor souls in Georgia who have been pummeled non-stop by Walker and Warnock ads for months. Here’s an idea: instead of spending $500M on ads, you are raising enough money to actually impact the issues you claim to care about? You could literally buy school supplies for every teacher in Georgia, or pay for police officer training or equipment – whatever your issue, you could impact it with that kind of money, and prove your ideas to voters!

University Endowments: Elite universities have fallen into the same trap as hospitals, in that they no longer spend significant sums toward the public good, but hide behind non-profit status as their endowments grow ever larger. Harvard, Yale, Stanford and other universities with massive endowments spend less than 4.5% of their endowment annually, while generating investment returns of over 8.5% annually over the last decade. They grow the cash pile further by soliciting donations that they have no plans to spend! Is the primary purpose of a university to be a brand for endowment fund-raising?

Honorable Mention: Donor Advised Funds enable individuals to “donate” money to a “charity” and gain a tax deduction, but do not require that the funds ever be distributed! As a result, $142 Billion in donations sat in DAFs as of mid-2021, generating fees for investment managers while not being used for any further societal purpose.

What’s the solution here? Has the American non-profit system become so bloated that it needs to be torn down? A simple starting step might be to raise and enforce requirements around actual charitable work – if private foundations are required to distribute 5% to charities annually, surely operating charities can be required to spend 5% of their revenue on bonafide charity work annually?

P.S. I think there are a great many causes worth supporting out there in this giving season. GiveWell and the Copenhagen Consensus do a good job researching both charities and areas where donations can provide maximum benefit. And you can see my family foundation’s grantees here.

Posted on by praveenghanta 0

The Future of Work is Here!

I’m excited to announce the launch of fraction.work on ProductHunt – upvote fraction.work there to help us gain exposure and change the future of work!

Readers of this blog know that I like to focus on big macro trends. The macro trend here is incontrovertible – working age populations are flat or dropping in every developed country on Earth. We keep hearing that the robots are coming, and that automation will take all the jobs – meanwhile US unemployment is back near all-time lows, despite a Federal Reserve moving rapidly to force a recession.

There’s only one solution: expand the labor supply. And the fastest way to do that is to tap into the millions of American workers willing to work more, or to keep working part-time.

At fraction.work it’s early days, as we are focused for the moment on fractional software developers. But in the software field alone, I estimate that there are 500,000 additional workers available on a fractional basis. McKinsey’s research shows that over half of all jobs can be done in a remote or hybrid fashion – fractional work opens the door to millions more employees filling open positions we can’t otherwise seem to fill.

Posted on by praveenghanta 0

To: Old-fashioned CTOs who think software development can’t be done part time

As I work to build my new startup fraction.work, I’ve come across the Availability Objection more than once. In essence, it’s some variation of “there’s no way a part-time developer could EVER be effective on MY team!”

In my latest post on the company blog, I outline how availability is a silly objection to fractional work for modern software development organizations. Of course if you’re still insisting all of your employees go into the office 5 days per week, perhaps you’re not modern enough to try this just yet…

Long story short, any CTO or VP of Engineering with a clue knows that half of a senior software developer’s time is worth many times that of most full-time junior developers (whose productivity is actually negative when they first start). So why wouldn’t you consider hiring fractional senior developers to help build your team out?

Posted on by praveenghanta 0

My experience as a fractional software developer

I started fraction.work earlier this summer, based on my experiences as a fractional software developer earlier in my career, and my experience hiring fractional developers while running HiddenLevers.

Those experiences guide me to believe that there’s a huge market for long-term, part-time software development work (that’s how I define “fractional” software development). We’ve seen fractional CFOs, CMOs, and GCs, but the adoption of this approach has been much slower at the individual contributor level and in particular in technology roles.

This is ironic because software development is more amenable to remote work than any other role – witness the explosive wave of offshore and nearshore development since the pandemic normalized remote work! Employers oddly feel more comfortable working with someone who half a world away and who may not grasp nuances of cultural difference, than working with someone in the US who is available 30 hours a week?

I know this isn’t really true – but many companies have a mental block when it comes to part-time work. As part of normalizing how effective it can be, I detailed the experience on the fraction.work blog – I hope you’ll follow the story there!

Posted on by praveenghanta 0

Uvalde + Buffalo + Parkland = Make guns 21+.

Uvalde, Buffalo, Parkland – the common thread in these massacres? All were committed by under-21 boys who purchased their guns legally. In mass shootings, 77% of the murderers obtained their weapons legally! Over 17% of all homicides are committed by those 18-20, and most weapons used in crime are obtained legally or through straw-man purchases from legal sellers [1].

So you’re telling me we could reduce school shootings and potentially stop 4000 deaths per year, just by making kids wait until they are drinking age to buy a gun [1]? It’s not quite that simple – with America awash in guns, eliminating access wouldn’t stop perpetrators entirely. But it’s worth noting that the aforementioned trio of shooters didn’t have criminal records, and didn’t have criminal contacts on whom to rely for illicit weapons. If only 1 in 4 young adults were stopped from obtaining a firearm, this would reduce deaths by over a thousand per year. From a gun-rights perspective, no right has been taken away – just shifted a few years to enable young minds to develop and gain impulse control (brain development actually ends at 25).

Most reasonable gun safety measures are supported by the majority of Americans, but this particular improvement was also enacted by a conservative state – Florida ended gun sales to the under-21 crowd after the Parkland shooting. If Florida can do it, then virtually every state politically to the left of FL should be able to make this change. Narrow Federal legislation in this regard might be possible (though unlikely) in the current moment [2]. As this latest tragedy focuses our attention on the issue, I hope politicians will focus on simple, attainable changes like these.

[1] The FBI data uses slightly different age ranges, but if we add 1/5th of the homicides committed by those 20-24 to homicides committed by older teenagers, we get 1910 homicides in 2019 – this is 17% of all homicides that year (where age of offender is known). When scaled to 2021 homicide levels (using 6.9 per 100k rate and Census 2021 population), this is 3893 homicides per year – 79% of which are estimated to be committed by firearms. That’s 3110 homicides per year. Using CDC data we find another 900 suicides by firearm within the 18-20 age group – for a total of 4000 deaths per year!

[2] Theoretically this should be easy to pass at the federal level, but Congress has become so ossified and reactionary that nothing will pass there.The guns-at-all-costs crowd has grown more extreme, with many calling for ALL weapons to be legal (yep that includes nuclear weapons, according to a former TX state representative).

Posted on by praveenghanta 0

Are COVID Deaths Changing The Electorate? Wisconsin Edition

I just wrote about a detailed calculation of COVID deaths and their potential electoral impact in Georgia, my home state. One factor that causes COVID deaths to have little political impact in Georgia: the state has a large black population which is disproportionately impacted by COVID, and this balances out deaths (politically speaking) among the older white population.

What about if we look at a swing state like Wisconsin, which is 87% white per the Census Bureau?

White deaths roughly approximate the white share of total population in Wisconsin.

This ought to mean that COVID would push the Wisconsin electorate leftward, correct?

Here are 2020 exit poll results for white voters in Wisconsin, and 2020 exit poll results for black voters in the Midwest (black Wisconsin voters weren’t available as a subset):

President Biden won Wisconsin precisely because he lost the older white vote by a relatively small margin.

Let’s do some simple math:

8064 deaths * 86.5% white = 6975 deaths

6975 deaths * (10% Trump margin amongst age groups at risk) = 698 net loss in Trump voters

8064 deaths * 7.7% black = 621 deaths

621 deaths * (62% Biden margin among black voters in Midwest) = 385 net loss in Biden voters

The Wisconsin GOP appears to have lost 300-350 net votes due to COVID thus far. It’s possible that this understates the impact, since the white voters that died post vaccine-era are increasingly represented by GOP voters (since they are more likely to refuse vaccination). But 80% of COVID deaths in Wisconsin occurred prior to general vaccine availability (prior to 1/31/21), lowering partisan effects due to vaccine hesitancy. Even if we assume a 20% Trump margin among white voters that died post 1/31, this only increases the GOP’s net vote loss to 500 votes (add 1/5th of the white vote * additional 10% margin).

The impact of a 500 vote swing could be meaningful in states where politics is a game of inches these days – but we can’t overstate it. Voters’ overall reaction to how the pandemic has been handled is by far the larger factor in how COVID impacts American politics.

Posted on by praveenghanta 0

Why Delta Might Be a Good Variant

I’m no epidemiologist, but it occurred to me that there might be a silver lining with the COVID-19 Delta variant.

  1. Delta is explosively contagious, with an R0 between 6 and 9. This means that the average infected individual is expected to transmit Delta to 6 and 9 additional people (versus 2.5 for COVID-19).
  2. This means that it just rips through populations. Because it can be transmitted by the vaccinated, it can travel deeper across the population as well.
  3. But so far is has not been found to be more lethal than the original, except to the extent that it crushes hospital capacity with its surge.
  4. Think about the virus’ evolutionary goals. It doesn’t actually care whether people live or die – the variant that spreads best dominates. If Delta manages that, while letting the vaccinated largely be unharmed, then it could become the pandemic’s endpoint.
  5. Why? Delta might crowd out new variants if it gets around and becomes the primary endemic version. To defeat Delta and take its dominant position, the next variant would have to be even better at spreading.
  6. Of course that is possible, but if Delta already can be spread by the vaccinated, it’s possible that is has already maximized the population growth prospects for this kind of virus in humans. The vaccinated can carry a Delta population in their nose without realizing it.
  7. Ironically, if any current vaccine were perfect, it would leave an attack surface for the next variant – conquer the vaccine. Perhaps the optimal vaccine simply allows us to live with the Coronavirus and welcome it to the family of standard household colds?
  8. And so perhaps Delta has evolved to be well suited to our current vaccines, able to maximize its reproduction. If we are lucky, it will crowd out any more lethal strains – marking the beginning of the end of the pandemic. There’s some indication that this might be happening, as Lambda has been present for some time, but has yet to grow in the US to extent that Delta has.
  9. This all assumes that vaccination efforts eventually meet with success and cover the overwhelming majority of the human population.

This could be wishful thinking – new variants will either validate or make a mockery of it soon. But the explosive growth and subsequent subsidence in COVID Delta cases in the UK and India give me room for hope!

Posted on by praveenghanta 0

Half of Police Homicides are Justified – A Data Analysis

Analysis of all 2019 US police homicides indicates that half are not justified – over 500 individuals per year die unnecessarily at the hands of police.

In 2019, police in the United States killed 1,099 people – and US police are tracking toward 1150 for all of 2020 [1]. While there is no uniform government database for police homicides in the United States, non-profit efforts like Mapping Police Violence have emerged to track the issue. While great work has been done collecting data, I’ve seen no analysis as to whether the homicides are justified. At one extreme, police unions argue that the police are always right – they believe that police homicides have a nearly 100% justified rate. BLM protesters and others argue the opposite – but where does the truth lie? If all police violence were justified, then there’s no reason for concern. As hundreds of videos and photos now show, it appears that the fraction is much lower – necessitating this analysis.

I analyzed fifty 2019 police homicides by hand, reading media reports, reviewing video evidence, and reading police reports. All 1,099 police homicides in 2019 were then analyzed using an automated approach – see the spreadsheet at bottom for the full details [2]. I used calendar year 2019 data, and manually scored 50 homicides using a list of rules as follows:

Rules Used in Manual Scoring: (51% of police homicides determined to be justified using these rules)

  1. Was the deceased provably (video, non-police witnesses) attacking officers or a victim with a firearm? If so, set to 100% justified
  2. Did the deceased kill anyone else prior to or during police intervention?
    If so, set to 100% to justified
  3. Was the deceased armed with a firearm or knife? If so, add 25% to the probability. (Cars, tools, and other implements are not counted here)
  4. According to the police, was the deceased threatening the police or a victim with a weapon? Is so, add 25% to the probability.
  5. According to non-police witnesses or footage, was the deceased threatening the police or a victim with a weapon? If so, add 25% to the probability
  6. Was the deceased shot in the back, while running away, or while driving away? If so, set the probability to 0%. (Shooting at drivers in cars has been proven to be extremely dangerous to the public and to officers, and is outlawed in many countries)

For the automated data analysis, I used only data available within the Mapping Police Violence spreadsheet.

Rules Used in Automated Scoring: (54% of police homicides determined to be justified using these rules, with all data per police reports)

  1. Was the deceased armed in any fashion? If so, add 25% to the probability.
  2. Was the alleged weapon a firearm? If so, add 25% to the probability.
  3. Was the deceased attacking the police or others at the moment they used lethal force? If so, add 25% to the probability.
  4. Was the deceased holding their ground and not fleeing? If so, add 25% to the probability.
  5. Was the deceased fleeing at the moment the police used lethal force, whether by car, foot, or other means? If so, subtract 25% from the probability.
  6. Did the deceased exhibit symptoms of mental illness? If so, subtract 25% from the probability.

Both analyses show that roughly half of all police homicides were found to be justified. When reading through and scoring individual homicides, I noted a wide range of cases ranging from truly heroic action to absurd and ridiculous [3]:

  • Heroic: Killing active assailants engaged in firing on officers or the public
  • Dubious: Shooting suspects in the back or in a car while they were trying to run away or drive away, even when they posed no threat
  • Absurd: A mentally ill person called 911 too many times, resulting in 911 dispatching officers to arrest him for excess calling, leading to his death unarmed and in his own home, after struggling with police

If half of all police homicides are not justified, then police are responsible for over 500 preventable deaths per year. This result cries out for change, even before potential racial inequities are studied! For those who think the police deserve the benefit of the doubt – the numbers indicate that the problem is real, and needs real attention. For those who think the police are always wrong – there are hundreds of instances in 2019 where the police rightly used lethal force. As usual in America these days, the solution is not binary – we need to acknowledge this and take reform seriously, but not to absurdity.

[1] Through August 24th 2020, policed had killed 751 people, according to Mapping Police Violence – that’s through the first 237 days of the year. Multiplying by 365 / 237 to normalize for a full year yields a rate of 1157 homicides per year for 2020 thus far.

[2] 2019 Police Violence Dataset with Justified ProbabilityDownload

[3] It’s important to note that the vast majority of the data for this analysis comes directly from the police. By 2019 anti-police violence protests movements had already gained traction across much of the country, leading police departments to proactively provide evidence when shootings are justified. When a police department refuses to comment or provide evidence on a shooting, the innocent-until-proven-guilty standard should be applied, meaning that the justification percentage is 0% in the absence of evidence.

Posted on by praveenghanta 0