It's near perfect idiomatic java. But does the dev solve any actual problem? More generally, how can you really evaluate a code project by parsing it line by line for syntax and style best practices? I think this algo would select the essays with the best handwriting, not the best ideas.
> Once a candidate links their online portfolios and projects—usually Github repositories—to Gradberry, the platform analyzes whether or not a candidate has written good-quality code.
Tell me one human being that can read someone else's code and determine whether it's "written good-quality code" and I'll have you a queue of companies that want to use you as a recruiter. AI instructions are created by humans. If a human can't determine the rules for "quality", then how is an AI going to do it?
The rules are inherently there; we humans just don't understand them other than as a case by case basis.
What needs to be done is a way to distill the logic behind "written good-quality code" and feed that to a machine to analyze. Machine learning seems like a good fit for this kind of exploratory problem.
The rules aren't inherently there because there is no one standard for good code. Code quality cannot be separated completely from the context in which it is written.
Not as a whole, but there are rules that can be evaluated against. Think consistent styling, usage of comments, etc. Things that the average linting tool tells you about.
Granted this doesn't prove correctness, or lack of bugs, but style is a major factor of readability and hence quality.
I agree style can be a useful input to evaluate. Most of the critics, myself included, take issue with the broad brush Gradberry uses to paint their ability to measure quality. It's masquerading subjectivity as objectivity.
Ok, so let's just pretend for a second "machine learning" solves the problem of determining whether code is good quality or not. Why not sell this as a service to analyzed large codebases for any software company? Don't you think every software organization in the world would pay thousands, if not millions, or dollars to analyze their codebase?
> Don't you think every software organization in the world would pay thousands, if not millions, or dollars to analyze their codebase?
Uh...no? If software organizations cared strongly about "high quality" code, then there wouldn't be so much terrible code and technical debt floating around in the world.
Companies care about:
1) Having code that works (for some value of 'works') to solve a problem, and
2) they care about getting that code immediately.
They only distantly, indirectly care about code quality as it hinders or helps #1 and #2.
Not sure about the approach to take a look in public code.
Students and graduates may still not developed good coding skills, assuming they have the time and energy to contribute to open source (they probably should as a pure marketing tool, though)
People with experience may not have their code available. Not every company works with open source.
It seems to be quite biased to certain kind of developer. I agree is a good kind of developer, but if the objective is to find "the best", it could be a relatively small talent pool.
Their number one tip is "Good profile pictures matter", so welcome back to prejudice land (I'm registering anyways, I'm a white dude, it shouldn't hurt me).
Hiring isn't just about how people code. Alot of times you run into other disconnected problems, like team members not getting along, people working in opposite directions, and so on.
A a Latino programmer, I must say it's never been easier to get a job. From this angle, there's something even beyond a meritocracy.
However, if one wants to raise money(some VCs) or get a finance job, most are biased towards a certain stereotype(skills not as blatant) and that's where the real prejudice exists.
I disagree. I had a startup in the recruting space. Companies only have about 3% to 7% of their jobs filled by head-hunter placement fees. This company is smart.
Most of existing head hunter placement fees were obtained because a head hunter getting a 20% to 33% placement fee was only able to be obtained by a lot of manual sales labor. This company automates that with a low 5% placement fee winning over competition and not requiring the high manual sales labor.
Second, this company is trying to find companies that will hire most of their engineers through their system, instead of only 5% of their company hires. Also placement fees at 20% to 33% are only paid on: a) exec positions, or b) when a position takes a very long time to fill. That small category limits the percent of employees able to be hired with a contingency fee (from 3% to 7%). This company being cheaper means that companies can hire all levels of engineers and do it for quality reasons, since the 5% is no longer a blocker.
5% is lower than the average recruitment fee, but it doesn't sound like this service is providing most of the same services (which are time consuming) that an agency provides.
This product scans code and sends high-ranking profiles to an employer. It appears as though that is all the product really does, other than take in feedback once a hire is/isn't made.
Agency recruiters are unlikely to be able to evaluate code. But an agency recruiter will review a resume, speak to or meet a candidate, send the resume and bio to the client, schedule interviews, debrief one or both parties after interviews, serve as a buffer in negotiations, perform reference checks, and close the deal. Those are the elements of recruiting that are time consuming.
5% seems low, but the product doesn't address many other areas of recruiting. If anything, I could see this product being marketed to agency recruiters as a way they can vet the code of the candidates they represent.
Yes, we pay ~30% of first year comp for traditional agency hires. Granted, they're doing more than simply flagging candidates of interest, but if this product is successful, I'm just as willing to pay 10% or 15% as I am 5%.
Maybe this is intended to be "disruptive." A lot of what is called disruptive, seems to do with replacing a labor intensive middleman with automation, at a lower but still profitable price.
I agree and was mostly observing that 10% also meets your last criterion.
Sell goods and services based on the value to the consumer (minus a consumer surplus), not on a cost-plus-markup model. The cost-plus model only needs to set the floor (so you don't run unprofitably forever.)
This company may be missing an opportunity to market this product to the industry that may benefit the most - agency recruiting (aka headhunters).
Most recruiters don't have the ability to evaluate code, so when they are representing candidates to their clients they are relying on a host of other indicators and data in order to try and gauge the coding ability of their candidates.
This product is a way for recruiters to vet the code of their candidates, and could then pass along those results ("Gradberry score of n") to their client as a value-add that other recruiters not using the system can't provide. A somewhat similar concept to recruiters years ago who asked candidates to take skills tests from places like Brainbench.
For the right price, I imagine recruiters would be interested in this as a differentiator between themselves and other firms. Gradberry would also still benefit from the feedback data if recruiters reported their results.
> According to Reuters, an overwhelming majority of the startups it looked at were founded by people who had held a senior position at a big technology firm, worked at a well-connected smaller one, started a successful company, or attended one of three universities—Stanford, Harvard, and Massachusetts Institute of Technology
Did they ever stop to think that maybe people who are talented and interested in starting their own tech startup business might seek out these schools and positions, disproportionately to other options?
You can't point at facts like these and claim they are proof for a lack of diversity. It might turn out that, given the chance to evaluate and interview every possible candidate in the world, that the top 1% of talent might be 80% drawn from those backgrounds, irrespective of all other traits.
Actually I don't see how being a manager at bigco or being a wealthy student in an expensive school with lots of means helps you bootstrap a company on a shoestring.
On a side note I'm more and more weary about the "lack of diversity" narrative, Indians and eastern Asians are over-represented, this is not a white world, there is a lack of Blacks, Hispanic and women, but remember that the Indians and eastern Asians will be the most suffering minority in rebalancing. Also, I think the issue of white women might not call for the same changes as the issue of minorities, because they are not a minority. I still don't know what to do (but I don't work, so you're free to take my job, I was a manager in IT), but I think a more nuanced analysis and a finer picture than "it's a bunch of white rapists" might help finding leverage points.
Thank you for actually pointing that out. As an Asian man, it annoys me that we also get excluded from being included as part of diversity. Asian men in tech? Who gives a shit, we need more x y and z! Not enough Asian men in Hollywood? Who gives a shit, they don't sell movies and the demographic of the U.S. isn't composed of Asians anyway!
> Actually I don't see how being a manager at bigco or being a wealthy student in an expensive school with lots of means helps you bootstrap a company on a shoestring.
I don't know either about the wealthy students, but being a manager at a bigco surely must help you to get customers with the network they have, especially if their startups are in a related field.
In my experience, it doesn't. They just go after blue chips again directly, but blue chips have a 18 months decision time on a purchase and don't like new technology. They lack the humility/will to go after many small companies to hone the product, business model and the funnel against many customers before hunting the whales again.
Actually I don't see how being a manager at bigco or being a wealthy student in an expensive school with lots of means helps you bootstrap a company on a shoestring.
I don't think the question is whether it helps you, the point is that it's a selection effect. Either the people who are smart and effective in those fields end up disproportionately wanting to go to big, brand-name schools and companies, those big brand-name schools and companies are able to hire those people, or both. It's possible that they also have a beneficial non-selection effect, but that's not what the parent comment was proposing.
What do you mean, 'did they stop to think'? The whole point of this article and their motivation is that's the thinking that already dominates a lot of the culture in the valley.
It goes something like this. Of the 7 billion people on the planet, anyone important will go to Stanford, Harvard, or MIT as a student. If they don't do that, they must prove themselves at an 'elite' company populated by such people so they can demonstrate their value and gain access to the exclusive funding supply.
Ironic as my first job was on campus at CIT Cranfield University which only did MSC and PHD courses and would consider those 3 as feeder schools
They used to say Cranfield considered Harvard, MIT , Ox-bridge as as paces where it was permissible to do ones "First Degree" as long as you did the right course and the right supervisors and got a First.
Is anyone else driven nuts by how hard it is to get at the data behind so many articles?
For instance, the bit about how it's important to have attended one of three universities - "Stanford, MIT, or Harvard", made me wonder what kind of disadvantage you are at if you attend a very strong but public CS program (Berkeley, University of Washington, Illinois).
The article has a link to the reuter's analysis here:
If you hover over some of the dots on the "unconnected" companies, you can see a few of the other universities attended. So it does appear that yes, attending Berkeley or other top public CS programs does put you at a significant disadvantage where it comes to getting series A funding [1], though this was as far as I could get. Maybe I'm bad at web searching, but I just couldn't find the raw data.
You know, as an aside… ycombinator set up an amazing series of talks from founders and CEOs for a truly exceptional class at stanford (https://news.ycombinator.com/item?id=8325479). There was a posting about it on HN a while back, and they did make the videos available… but there's nothing like actually being there, possibly with the opportunity to meet a high up in person.
There were a few objections, some down-modded (for a variety of reasons), about possible class issues here. I would say this: there's a reasonably good CS program across the bay at a university that enrolls more low income students than the entire ivy league combined. If middle and low income students appear to be having more trouble becoming founders and getting funding (as opposed to becoming high value employees for those who get series A funding), this might also be a great location for another round of this course.
[1] this is highly relevant, since students from top public CS programs don't appear to have significantly lower mid-career salaries as workers (in fact, worker bee salaries may be slightly higher out of Berkeley).
Also I may have made a correlation causation error. Could be that the university itself has little to do with this.
Silicon Valley really, really wants to be a meritocracy.
Nope. It really doesn't. It wants to be perceived as a meritocracy. That's a different goal entirely. It's the same privileged, well-connected "MBA culture", but in the 21st century, MBA stands for "Meritocracy By Assertion".
VC doesn't want not to be a clubby, relationship-based business, because who-you-know cultures are great at giving the well-connected an extortive power that they wouldn't otherwise have. ("Those who question the meritocracy will be de-meritized.") Sure, there are Silicon Valley engineers who want to live in (and believe in) meritocracy, but they're not the people with the power. You can get an employee position on merit, but if you want to ascend to the founder or investor ranks and get real equity in the Valley... you still have to be a child of the old empire.
It was hard to read the rest of this article after this critical miss.
https://github.com/EnterpriseQualityCoding/FizzBuzzEnterpris...
It's near perfect idiomatic java. But does the dev solve any actual problem? More generally, how can you really evaluate a code project by parsing it line by line for syntax and style best practices? I think this algo would select the essays with the best handwriting, not the best ideas.