I got frustrated with ChatGPT for hallucinating sometimes, and my friend said "could be worse, could hallucinate all the time." And I thought it'd be fun to make that, and then we came up with a few more, and now we have this: https://cap-gpt.onrender.com/
Ask HN: What is the most useless project you have worked on?
201–210 of 794 posts
Re: Ask HN: What is the most useless project you have worked on?
#202Earlier quoted context omitted.
Awesome story, and a good lesson in "always get it in writing"
The concept seems weird to me. I have never been given a mark by a teacher that matters. Exams were administered by independent bodies.
In the UK that’s the case through high school and college (UK definition, 16-18) but at university the exams and coursework are graded by the lecturer/professor. So that’s likely what happened here.
Re: Ask HN: What is the most useless project you have worked on?
#203Whenever people talk about replacing engineers with AI, I think about situations like this.
When the first fully machine automated software consultancies post their first cash flow statements, there will still be humans building features that will never be used somewhere else in the same industry.
Re: Ask HN: What is the most useless project you have worked on?
#204Earlier quoted context omitted.
Wow, I once added clippy to a website I worked on too :) It wasn't an all-nighter, though! I used clippy.js, so it was just a few lines!
Ahh that was it, yes it was clippy.js! It was a desktop app though (with a C++ & Python API) that we had built a lot of integrations around, so there were a lot of Qt schenanigans going on, and I think he had attempted to get it to actually try and search some internal DB too for witty remarks.
Yep, adding to a non-web app is definitely harder!
Re: Ask HN: What is the most useless project you have worked on?
#205I worked on a six-month contract project for a large US retailer; the fact that everybody else I worked with was laid off the second week probably should have told me something was wrong, but they kept paying me so I kept showing up.
Anyway, the only task they ever gave me was to build Cisco router ACLs to match existing traffic in the stores; they needed to implement access control but wanted to make sure they weren't blocking anything important. Because many of the stores had been opened years before, there wasn't necessarily a consistent IT stack in the store; there were a lot of one-off solutions so they wanted a universal set of rules with per-store exception lists.
So, every week, someone would drop a few terabytes of network logs in an FTP server, separated by store, and I would distill them down to a set of rules; if something was being used in more than X number of stores, it went in the universal rule list, if it wasn't it went in a store exception list. At first I was doing it semi-manually, but eventually I built a database and wrote some SQL to mostly automate it - got it down to about two hours a week, most of it waiting for the DB, by the end. Once it was done, I would send all of my updates to a network engineer who was the designated point of contact for the access control project.
They had a lot of stores, and for some reason could only supply a certain amount of logs weekly, so when my contract was up there were still a number remaining. I tried to set up a meeting with the network engineer to go over the work to be done and the automation I'd built, but he never responded. Eventually I tracked down his desk in the vast corporate complex and paid him a visit. He was pleasant but told me that he wasn't even on the access control project and he thought it had probably been canceled at some point. He had been dutifully copying my updates to a network share somewhere in case they were ever needed. I gave him the SQL scripts and the database info and he put them out on the share where they probably still are today, a decade and a half later.
So that was six months of my working life.
Re: Ask HN: What is the most useless project you have worked on?
#206I had a boss at one of my last gigs who would cook up the most impossibly over-engineered (and badly designed and rarely worked) projects ever.
The worst one ever was he wanted to build a kubernetes observability platform from scratch. Specifically, we wanted to filter our application logs for errors and be able to find them. There's a million out of the box ways to do this with free and enterprise tooling, like EFK/ELK setups, datadog, etc. But no. We wanted to do this from scratch.
Well, fine I thought. Good resume builder. His approach was roughly this from what I recall:
- On an EC2 instance we will set up permissions so it can access the control plane of our entire global EKS infrastructure (like 4 dozen large clusters hosting 10k+ containers)
- We will run a series of bash scripts in a background process that hits the control plane every few seconds with Kubectl to get raw pod logs and store them for analysis/error grepping (probably someone is laughing already)
- We would use a similar series of bash scripts to automatically generate a mostly static website that would link to the paths of log files deemed "problematic" in some insane filepath system that was approximately organized by timestamp, the idea being if you thought an incident happened in Cluster XYZ at 11am you'd navigate through this web of raw log files and find the directory that had the precise timestamp you needed (linux epoch of course, not mm/dd/YY)
- Because new files were generated all the time, like every few seconds, we also had to periodically "refresh" the site by rebuilding the static site with all the new links
All told I think it ended up being some absolutely psychotic mess of 50+ scripts and over 15,000 lines of almost all bash and some Go templating.
Surprisingly, it did what it was supposed to, it just looked terrible and had an obviously bad UX. Luckily before we could show it to anyone important the exact thing happened I had warned about and we started crashing some of the clusters' control planes from the sheer number of requests we were sending to it. He panicked and told me to shut it down.
I'm sure I don't even need to go into detail how useless this is. I did become a wizard with kubectl, however, so it was valuable to me. Not so much to the company.
Re: Ask HN: What is the most useless project you have worked on?
#207Earlier quoted context omitted.
I’ve been playing with my son, but using my shoebox full of old cards, many with an “IV” logo so must have been around 4e that I was playing back in the day. The game is complex, but a good intellectual challenge for a bright 6yo reader.
Probably VI; 4th edition didn't have a set symbol.
Right.
Also, FWIW, 5th edition also didn't have a set symbol, but the "Visions" set was released at around the same time and had a "V" symbol... (I only realized how confusing it was after 27 years lol...
Re: Ask HN: What is the most useless project you have worked on?
#208Some years ago, a colleague literally spent all night in the office, on the 31st of March, implementing clippy for another third party commercial application (not MS Office) that was heavily used by the company and across that industry. He inspired me to take April Fool's seriously. I've since changed industries but last year I jumped on the AI bandwagon with https://chatellite.space and this year a nod to a popular…
Really cool! chatellite.space
Re: Ask HN: What is the most useless project you have worked on?
#209I worked for a company that used to resell a single-board computer used for control applications. The vendor was getting antsy because we were always late paying our bills. Our company owner didn't like their complaints, so I was told to design one that was "better than theirs." That's it. That was the entire spec. OK, I was also told that customer X would like a board that they can program in C (current board had a…
> That was the entire spec
To be fair, that's how most ideas start off. Your boss gave to free reign over a new project. You should have been happier than you seemed.
Re: Ask HN: What is the most useless project you have worked on?
#210As long as you have been paid for that "useless" project, you should be happy about it. You learned something...plus..you can showcase that "useless" project on your profile or something to get new work.