I failed a take-home assignment from Kagi Search

My honest feedback below from the perspective of a hirer. I'll start by saying I hate takehome stuff, for exactly reasons like this. It wastes everyone's time. They're fine as a 'last step before hire' thing, but not as a filter.

1 - Too much chatter. Part of the assignment is using judgment and working in ambiguity. I probably would have ran with what was given enough to knock out something small and local in an evening or two. Asking questions is usually fine, they even welcomed it, but seems counter to the original ask.

2 - Writing and sharing a proposal seemed like way overkill. You have to remember that these companies are now getting hundreds if not thousands if not tens of thousands of applicants, that is a lot to deal with if everyone does so. I think it's a bit of a disconnect...you feel like you're going above and beyond and being thorough, they feel like it's being a bit long winded and wasting time. That probably explains the nonresponse.

3 - The finished product seemed functional, but seemed a bit overkill on the infra and polish. This is probably a good thing to work with you, but ended up wasting a lot of your time if not being selected, which was the case.

4 - Maybe I missed something, but the requirement asked for terminal inspired. I'm not quite sure precisely what they meant by that, but didn't see any possible interpretation of that in the result.

Anyways, hope you don't take it too negatively or personally - you obviously are a talented individual and moreso seem to really care about your work. Just wanted to play a little devil's advocate with a different perspective.

  This project tests your ability not only to code, but to deal with ambiguity and open-endedness

My first though on looking at the code, and then on demo video was: wow, it took that person a solid week to write a web app with two pages.. all while missing the most basic email features, like an ability to opening a message (and not just show full text of mail bodies on one page).

Later on I read the requirements... This person applied to a position named "Email Backend Engineer" but they actually used a third-party product (postmark + turso) for email backend! They also clearly don't think about email too much - the basic stuff, like plaintext email formatting, viewing, and folders (at least inbox + outbox) are simply missing; while there are optional features, like login screen, admin page and backend framework. And backend database does not even containing a email headers map!

That person might be a great engineer, but I don't think they would be a good fit for that specific position. I'd reject them as well.

(A separate question is that hiring manager's second response... We don't do take-home interviews, but I imagine I'd be stumped by a proposal like that, it seems so irregular in a take-home assignment. Still, I can see how the candidate took that response for an approval... perhaps a next candidate would get an extra sentence perhaps something like "actual problem grading would depend on the code quality, number of features implemented, and how close those are to requirements and job description")

Update: read original job description and not just the except from the blog. That person surely omitted some stuff in their blog! the original title was:

"The project is to build a minimal, terminal-inspired email client"

it gave examples:

"Your implementation should take inspiration from existing terminal email tools like aerc, mutt, or even something like himalaya."

wow! that's 100% no hire, serious inability to read the requirements.

Take home assessments can be a valuable part of an interview process but they absolutely need a time limit. I think 2-3 hours is going to give you all the information you need, unless what you're selecting for is new grads with no dependents, hobbies, or responsibilities.

If this had been limited to 3 hours then the worst case is that the candidate would have lost 3 hours, but far more likely is that they would have come up with an entirely different proposal and/or solution that was appropriate for that timeline, and that extra information would have made it clearer what the company was looking for.

The other point I'd always encourage applicants to confirm is: are you looking for any answer, or are you looking for a good answer. Some take-home tests are purely about passing a test suite and how you manage it doesn't matter, some would prefer that you meet 80% of the requirements but write better code. I've seen applicants do the wrong thing on both sides of this.

I understand that their responses were very minimal and not helpful, but it clearly says in the requirements to make a "terminal-inspired" email client. However, in the video you shared [1] I see a somewhat generic email web app, with nothing particularly "terminal-inspired". Considering the fact that they wanted either a real TUI or a webapp, I'm pretty sure that the web app should've also been "terminal-inspired". Am I missing something? I'm not trying to necessarily criticize you, perhaps I just misunderstood the requirements.

Edit: I've actually checked the full requirements page, and they explicitly say this in "Inspiration":

> Your implementation should take inspiration from existing terminal email tools like aerc, mutt, or even something like himalaya.

[1] https://www.youtube.com/watch?v=yY1sVXMkP_o

As an engineering interviewer I feel the urge to comment. I like neither leet code, nor home assignments. They are both time consuming and provide you with a little information.

But having said this I would have likely rejected the author too. Kagi Search is a startup, these folks are typically looking for fast moving, pragmatic optimists who can strike the breadth - depth balance.

We used to have a colleague. They would collect the inputs, close themselves for a few days, come up with a solution only to learn that reqs changed in the meantime. It was not pleasant experience for anyone.

I took a look at the code, to see how I would react as a reviewer.

In the first file I opened, I got as far as here: https://github.com/Sleepful/mymail/blob/main/app/router/page...

  // Create a context variable that inherits from a parent, and sets the value "test".
  // Create a context key for the theme.
  ctx, err := getEmails(pb, e)

First line is very weird, unrelated to the task. Maybe copy-paste from a sample in a blog post? Anyway not paying attention to leave it in.

Wording in the second line is not consistent with third.

I’d stop my review there. Lack of attention to detail. Author does not demonstrate an ability to think clearly.

I did a recent round of interviews and this experience closely mirrors mine (at multiple places). Delivered an exceptional solution to the problem only to be rejected without a discussion about the delivered project. I have been on the hiring end of take-homes multiple times and firmly believe that if you ask someone to do a take-home assignment you must follow up with a chat about the code. Apparently that's not the case at most places.

If you're asked to do a take home, I highly recommend confirming that there will be a follow up regardless of the assessment, and if they do not agree to this, do not do the "home work". The honest truth is that most of the teams hiring are of pretty low quality and therefore implementing a good solution is a negative because the team hiring is not at your level (which is a frustrating reason for being rejected).

I'm an early adopter of Kagi and am now planning on cancelling my account over of this. Completely unacceptable. If you don't have time to chat with a candidate about their work, don't ask them to do the work in the first place.

This is a classic case of misreading the subtext. They want you to be independent and able to create your own work and goals. The ambiguity of the take home task is not there for you to interrogate over multiple emails, rather its an open palette for you to show how you take a relatively broad task and produce your own interrogation of the problem space.

As someone who works at a university this is very reminiscent of students who don't understand the assignment then complain when they get an unexpected mark.

I worked with a lot of engineers in my 40+ year engineering career. One thing that most engineers don't like is ambiguity. I didn't like it either, until mid career when I was thrown into an environment where the customers didn't really know what they wanted, and it was impossible to explore all of the solution space within time and budget constraints. I ended up thriving in this environment, but I watched as many qualified and experienced engineers floundered and over-stressed themselves. The attrition rate in this environment was about 80%. To survive, I had to use intuition, speculation, and innovation -- things I had not used much in my previous engineering roles. I sadly watched while a lot of really good engineers chose to move on, or were assigned to less relevant roles.

In this case, I perceive that the author (Jose) was looking for firm requirements, and did what he could to elicit them, but the Kagi folks were just looking for innovation and a demonstration of competency. They didn't care as much about the details of the submittal as they did about seeing what the candidate would do without specific requirements.

I just can’t get past the fact that he didn’t even deliver on the core deliverable of “ The project is to build a minimal, terminal-inspired email client”. Nothing about the demo I watched on YouTube looked “terminal-inspired”

Man, I hate take-homes.

When I see instructions that say “This project tests your ability not only to code, but to deal with ambiguity and open-endedness”, it actually means either a) this exercise has been so poorly conceived that we didn’t bother with a rubric or b) the work environment is so chaotic, we never clearly define specs or requirements for anything.

That, and the instruction to simultaneously “show off your skills” and deliver a pragmatic functional solution are completely at odds. Good simple solutions aren’t flashy.

I have done this dance a few times. My learning about interview assignments is review the written grading rubric before reading anything else. If the grading criteria lacks specificity then don’t do the assignment. If there is no provided grading rubric in advance then don’t do the assessment.

I feel like Kagi was just asking "Impress us" and OP misunderstood the assignment by actually building a solid project and handling edge cases that no one cared about.

Anyways thanks for writing this up OP, it was an interesting read and I am a Kagi customer so I liked learning about them.

Man this sounds rough. I had a somewhat similar experience a while back but instead of going back and forth trying to nail down requirements, I went down this spiral of wondering if the lack of clarity wwas part of some kind of meta-test.

"Do they want to see if I can build something with as little guidance as possible?"

"Do they want me to push for more requirements like I would on a real project?"

"If I build something cool but totally off from what they expected will that make me look better or worse?"

"Or are they just trying to weed out the people that can't code at all?"

In the end I didn't get a follow-up interview and they refused to give me any feedback on how I could have done better on the take-home assessment.

Back to the OP: One such example would be to do a live code review. This could be done asynchronously or synchronously. It could allow actually talking through topics and issues that relate to the challenges in a real software project. This would allow to surface much more of the knowledge in an experienced software engineer.

I like this idea a lot.

Building a concrete, working, minimum-viable solution from ambiguous requirements: that's the point of the exercise. That's what hiring managers want in candidates because those candidates end up being good at building a concrete, working solutions from ambiguous requirements. Which is every software project ever. Although the AI Age has unquestionably changed the efficacy of this kind of candidate screening, that is orthogonal to this discussion. For many years it has been one of the most-effective ways to screen for the ability to build concrete, working solutions from ambiguous requirements. Which is every software project ever. So it's no surprise it persists.

It's a tough lesson, but a good one to learn early: never waste time with interview take home assignments. This is especially true for broad and ambiguous assignments that you will need to sink lots of time into.

What if there are 300 other applicants? What's to disincentivize giving all of them the assignment even when there's only one open position? There is no guarantee that anybody will even clone your repository.

In a live coding interview, the company is committing valuable engineering resources to evaluate you. You have some assurances that they actually want to hire someone and not just get free labor or waste your time.

I understand the frustration here. I assume Kagi has a lot of candidates and is trying to identify the best ones for handling ambiguity. While I can appreciate the interviewee’s desire for more information or a rubric, it might potentially hurt the effectiveness of the take-home assignment.

During my college days, I had an on-campus interview with Microsoft that served as an assessment for whether I would be eligible for more in-person interviews in Seattle. The interview question was open-ended: “I would like you to design an alarm clock.” It seemed like an unusual open-ended question that prompted me to consider several critical factors, such as the intended audience, physicality, visibility, and audibility. By considering these factors and not immediately presenting a solution, the interview team informed me that I had passed the test. If they had provided me with more specific instructions or guidelines before the interview, they might have missed a signal that was important to them.

Unfortunately, I did not advance to the subsequent rounds at Microsoft. I interviewed during the fall semester. While I made it thru the full process, I was informed that I lacked confidence, but wasn’t a no. I was encouraged to reapply in the spring. The initial experience had left me drained and disheartened, so I decided not to pursue a second attempt. I now deeply regret this decision and wish I had given it another chance. If anyone is in a similar situation during their college years, I strongly advise them to avoid making my mistake. Working at a FAANG company could have changed my entire life and career trajectory.

I don't mind take home assignments per se, but it seems like there should be a reasonable "this should take this long" time boxing of the thing so people don't waste a ton of time. I've done take-home things a couple of times and they always had "this should take 12 hours" or whatever time amount.

The "simpler" solution part stood out to me though. Given that Kagi was putting little effort into the thing, I wonder if his verbosity/try-hardness might have been a turnoff? Afterall, part of the thing is determining if you want to work with a person, my impression reading his blog post was.. mildly turned off? There was just a bit of desperation to it (which might be from real factors!) but still. That sounds mean and I don't like saying it, that's just kind of the vibe that hit me.

You had the right idea to begin with - never do take home assignments.

Second - this is the wrong approach. Nothing wrong with writing a small doc. Write it, then implement the necessary features (terminal inspired client that can read and send email), and test it. Literally a cli for read and send, tack on curses or use imgui. Do IMAP/POP/SMTP or emulate it. The job position is for a backend role in the email team.

Tack on bells and whistles later.

It sounds like Kagi asked for something simple but creative, and the author overengineered it.

The author didn't really ask for a code review, though a lot of people imagining themselves in the position of technical hiring manager are already taking care of that.

Anyway: I don't think homework assignments are a valuable mechanism for filtering out candidates. Setting aside shoulds and shouldnots and the ethics of it and everything else, it simply provides a really poor signal for the value of a potential candidate.

Especially today, when something like this particular assignment could just be pasted into any of the popular LLMs and returned as completed within a couple hours.

If an organization wants to hire developers that can take vague project descriptions and convert them into code that may or may not do whatever was in the heads of the people that wrote the project description, then (a) that's a bad practice, but (b) it would be better for everyone involved if you handled this over a video call. If your staff are too busy to do a small pile of video calls with potential candidates over a couple of weeks, then they are too busy to properly onboard a new candidate and you should probably just pack it in. (i.e., they are not too busy to do this.)

If the goal is to hire somebody that knows IMAP, SMTP, POP, etc. inside-and-out and can crank out RFC-compliant code without using any of the third-party libraries that already exist for this sort of thing, then the assignment description should have asked for that.

Homework assignments are an unfortunately common practice, but worst of all, most of them are carelessly designed.

Take home assignments I feel have the second worst signal for any hiring workflow (behind automated leetcode), you give someone free reign to spend too much time working on something useless without seeing how they actually think or approach problems, you only get a solution at the end, and the solution, especially in open ended problems like these has very loose objective measures. If you want a long term hiring process, I'd just stick to a contract project, though personally just do monitored interviews; I know it takes up more of the hiring managers time, but the amount of signal you get from actually talking and interfacing with a dev is pretty useful

You probably triggered LLM blindness with your proposal. Never use formal style speech in markdown with bullet points.

I went through something very similar — spent several days on a take-home assignment, took it seriously, carefully followed all the instructions and delivered everything.

Then I waited for two weeks, and all I got was a clearly templated response.

The worst part isn’t even the rejection — it’s the vague replies, or no reply at all.It just makes you feel like your time doesn’t matter.

Thank you for writing this. We really do need more people to speak up about these things.

Very insightful idea here. If you work in software, requirements engineering is a key task. Figuring out “would this deliver for you?” is a key question you hash out with clients.

But the absurdity of take home work like this is the company feels obliged to keep their requirements secret. Thus, by doing the task not knowing if it will be up to spec, you compromise on your most important skill!

From this post, it seems like the hiring manager is either not the real hiring manager (maybe just an HR agent with this lofty title) or was himself hired without any screening or on the job evaluation for reading and responding to emails (which is a big part, IMO, of any manager) or has multiple roles to play and has no time to devote to hiring.

I feel sorry for the author to have spent more time on actually doing the assignment after that reply.

I did one once for Barclays for a Haskell role in London. A whole evening of work and they never even sent me a rejection email.

It’s pretty disrespectful in my opinion. It would be better if they took 90% of their applications and secretly threw them in the garbage than to waste applicants’ time evaluating them on vague criteria.

I have been lucky enough to be able to reject take homes. Next time, I might offer to do them for a flat rate of $200/hour but working for free is something I will avoid.

If it’s going to be used as a screening process, at least give clear guidelines and evaluate it pass/fail.

The requirements for even senior software engineers now are absurd (to say nothing of entry/mid level, which I think may not even exist anymore at most shops).

Companies are abusing their position and we should remember the most egregious of them and avoid them in the future.

This (and other terrible interviewing processes) seem like the epitome of a problem being tackled from the wrong end: Trying to filter things coming OUT of the hiring funnel instead of filtering the things going IN to the hiring funnel.

Case-in-point: Modern job sites (cough LinkedIn). Every job listing has HUNDREDS of applicants within an hour of the job being posted. It's ridiculous. It should not be _that_ easy to apply for a job.

The outcome is what we are seeing today: A company posts a job, is inundated with 100s/1,000s of applications. In order to filter out the 80% of applicants who aren't deeply interested in the role, the company deliberately assigns busywork/road blocks to slow down the process.

The other 20% of applicants will then spend days/weeks/months in the hiring process on intro videos, take home challenges, etc. Basically anything that can be throw at the applicant that isn't time with a human.

The takeaway for each can be broken down into:

- 80% of applicants: didn't spend anything, didn't get anything, don't care

- 19% of applicants: spent time doing some/all of the busywork, _aren't_ hired, end up very frustrated at the amount of time/energy/resources that was spent only to be discarded

- The 1%: spent time doing some/all of the busywork, _are_ hired, feel great!

I'm not sure what the exact solution is, but I know that:

a) it's a race to the bottom (with the bottom being full automation on BOTH sides of the hiring process),

b) I'd much rather spend a fair bit of time putting together applications for 5 jobs and be seriously considered than spend very little time on 50 jobs only to be immediately rejected or handed an assignment before I even talk to a real human being

c) hiring teams hate the status quo, too.

> We receive a lot of interest and applications for each position at Kagi which makes selection processes is extremely competitive.

This is the real take away. Don't put in outsized effort at companies that are highly subscribed. Everyone prides themselves on being "extremely competitive" and unless it's a FAANG (or well known quiet money fountain like Valve), it's probably not actually worth the effort.

If it's hot on HN, it's highly subscribed -- you just can't expect effort to be valued the same way when there are 10 applicants.

> We normally don’t provide feedback at this stage. We have had other submissions that were simpler and stronger, so we decided to continue forward with these candidates.

This would have been a great place to push and ask for a little more feedback (while being nice as possible since they have nearly no incentive to share more)... How could the other solutions have been simpler and "stronger" (whatever that means? more robust?) -- a few details on what you missed that they were looking for (tests? documentation? CI?) could at least add some value.

All that said -- finding a job is really hard right now. Wish you the best.

This is the same thing that happened to me with an Expensify interview back in 2011. After that, I decided to only do interviews in person, or where the person was on a call with me. Take home tests are not worth the risk of this crap.

Is it just me or he missed the point completely?

Look at all the third party services he uses! Nobody sane would pick that solution. That would be a completely maintainment nightmare.

He did not write an email client, he wrote a wrapper to some email client, which handles all the heavy lifting.

At the very least, write something that interact with a real (or fake) SMTP server. Other things are just cherry on top.

I feel Kagi's entire point of the take home assignment is to see how people handled assignments if left to their own devices with little instructions to go off.

When good companies are trying to hire engineers, they don't hire based on skill alone. They also try to hire based on how a person performs given certain constraints like time and ambiguity.

Sharing a proposal and asking for more time was probably the polar opposite what they were looking for

I think a lot of product managers prefer to see simple, fast solutions to a take home assignment. Understandable and turned in on time.

I read so many responses (the replies here to this post and to so many other similar posts) of the form

"Oh, I simply refuse to do interviews [of format xyz]. It's just not worth my time..."

I feel like such a worthless piece of crap when I read these. I apply to hundreds of places, and when someone comes along that even wants to talk to me or a miracle happens and I get to any sort of technical round, I simply DON'T HAVE THE OPTION to be turning them away for any reason.

Like, these companies have all the leverage right now, and lots of us have none. Responses like "just tell them no" seem so tone deaf. I guess some of you must be seriously baller, but it seems otherworldly to me.

This is close to how I feel when I read posts on Linkedin "demanding" that companies dispense with some unethical hiring practice (e.g. keeping up job postings for which they are not hiring, rejecting candidates without feedback, endless interview rounds, etc). I'm like, "Dude! The people who are doing this do not give a rat's ass about your preachy Linkedin post. We have no leverage here. Spare me your clickbait."

End of rant.

Without wanting to be unkind, this author misunderstands the purpose of the exercise, which is to demonstrate to another human being that you can write code that meets the company's standards (as in, would pass PR review), not to ship a fully-functional product solo. The initial email makes it explicit that they don't care if the product even works ("Can use a fake backend") and that the specifications are deliberately open-ended. A little intuition and empathy can tell you that they probably are not going to spend many hours reviewing a complex submission, so the project should optimize for demonstrability of code quality, not completeness.

"I would like to know what kind of response I could expect..." - This is also established from the beginning: you can expect either a "Looks good, let's do some interviews" or a "Sorry, not interested," based on the code you submit. They can't narrow down the choices prior to your submission, because they're grading your submission, not your proposal document with an extensive list of details that they already told you they're mostly ambivalent to.

"So it is funny that my project is so weak, yet it made them update the guidelines to something stricter." - Main character syndrome. As someone who has been on the other side of these kinds of reviews, the far more likely explanation is that they kept getting submissions where the build instructions didn't work (which is not disqualifying by itself; the authors may not be on the same OS as the reviewers) and they got tired of spending time dealing with it.

Ultimately, the failure here was not a technical one but a social one. The author tried very hard to do the thing that it seemed like they were asking for, not the thing they actually wanted. The hiring manager's unwillingness to engage with the proposal doc was itself a form of communication that they were not interested in this level of detail, it was just an implicit one.

It's common for engineer types to want all that kind of communication to be explicit, and I have a lot of sympathy for those folks, but the reality is that teamwork is a skill, and the ability to suss out what a stakeholder actually wants, but isn't saying due to incomplete information and/or office politics, is a reasonable thing to select for. The ambiguity of the prompt is a feature, not a bug: it's kept the author away from a company whose communication style they're not compatible with.

(that said, this project does sound excessively complex for a take-home)

I love Kagi as a product. I do not love Kagi as a company (based on reports of founder, posted here: https://news.ycombinator.com/item?id=40011314)

I think this is a classic case of over delivering.

The requirements to spin up and obtain live services (postmark, torso), is a bit much, especially for a take home assignment.

Personally, If your app can’t be spun up and tested with a simple "docker compose up —build", then you are doing something wrong.

My rule of thumb for take home assignments: if I spend more than a few hours on it. Then I am probably over delivering and need to rethink approach and remove shit I don’t need. There’s no way in hell I’m spending a full working day, let alone a _week_ of free labor for these assignments.

Note: also not a fan of "take home assignments" either. Of the few companies that ask for these, I consequently ask to be compensated for my time. Only 1 company agreed to compensation, but others declined and I moved on.

did the take home assignment precede a chat with the team?

I tell everyone up front (recruiters and companies) I absolutely don't do homework assignments as part of a job interview. If they're unable to judge my skill level from an interview, that speaks volumes about them, not me.

Furthermore, if they are even the kind of company that thinks they're so special they can or should even ask for that, then it proves they're the kind of people I want to avoid like the plague to begin with. So when I'm asked to do a homework assignment it's actually a huge time saver for me, because it tells me right up front to avoid them.

All developers should refuse this. But there are always enough that are so desperate they'll do pretty much anything.

The (interview) game has changed and it will get worse because of more layoffs this year.

To standout, you have to be a bit creative and you must sustain yourself with your own company / startup. (4 basic instructions here) [0]

Companies do not care about you or your take home solution(s) unless you've built something that threatens their existence with a competing product that makes money and steals their users.

Stop playing by their rules with these interview 'games' unless you have lots of time. Spend your time elsewhere. You now have AI, and you should build a startup instead.

[0] https://news.ycombinator.com/item?id=43212438

just use cursor lol

Yikes, I love Kagi but if I'd applied for this position and they gave me this assignment, I'd have told them to take a hike. This isn't an afternoon project, it's a full weekend. Maybe two. And are you supposed to be writing tests and stuff for this?

Moreover, this exercise tells you so little about the candidate and their experience. Sure, you know they can code (or did they just use Cursor?) but you have no idea whether they are good and addressing prompts and not at making wise choices. Who cares if the program can send email? What matters is that the client doesn't choke after 10k messages are received or that unicode is handled well: that's the sort of actual challenges you'd be doing at a company like Kagi, not making enormous toy projects.

Edit: Despite that, it seems like a bold choice for the author to build a web app instead of a TUI like the instructions lay out. One could say "terminal inspired" could include web applications but I think that's a stretch.

Honestly, you lost it with your first email asking for more direction, and every email after that just made it worse.

This was part of the assignment:

> This project tests your ability not only to code, but to deal with ambiguity and open-endedness, which is essential in R&D projects like Kagi Labs.

It was vague on purpose, they wanted to see how your decision making process goes with what you fill in the blanks with, without needing hand holding. You seemed to ignore this critical piece and went further and further against it with each message.

“but to deal with ambiguity and open-endedness, which is essential in R&D projects like Kagi Labs”…. Also based on the nonsense responses by the hiring manager, it sounds like one of 2 things

Either someone had a vision and is saying ‘Read my mind’

The alternative explanation seems to be, there is no vision, and the interviewee needs to define it.

In either scenario, the amount of communication, feedback, specificity, lack of respect for the power dynamics is appalling.