Last week, I shared a link on Hacker News about my startup and it unexpectedly soared to the top for a little while, with now 81 points. This is a post-mortem on what I learned from that exciting (and nerve-wracking) experience.
My product is an AI language teacher: you learn from speaking into the microphone and listening. I want to make the joy of learning a language through speech available to everyone, because learning from text gets dull and results in a thick accent that's hard to fix.
Stats
After a failed launched on Product Hunt I was apprehensive to post on Hacker News. But it resulted in more traffic in a day than I’d get in a month:
What I did right
A little context: I’m trying to follow MAKE by Pieter Levels. A few months ago, I worked for 10 days straight to prepare for a Product Hunt launch. The result? No one really cared. But I was hitting my third burnout in a year.
- I was apprehensive to share on Hacker News because it has a reputation for being an honest but tough crowd. I fought the apprehension by arguing with ChatGPT (our conversation) about what to post and how to post it, trying to use models as inspiration. It helped me iterate to be more structured, succinct and (hopefully) relevant.
- The posting time was probably right. I posted at 8 am PST (5 pm CET), which means both the American and European crowds had the chance to check it at the same time.
- This is still an on-going process, but I’m trying to fix every issue addressed in the comments. I’m maybe one fourth of the way into it. Good, because I’d rather implement a lot of small fixes now than overthink and procrastinate forever.
- Putting a signature link to Twitter on every page to facilitate contact – it did lead to nice contacts.
What I did wrong
featuritis
Gliglish includes a feature I really like: multilingual speech recognition. It lets you mix languages to ask things such as “What’s the difference between voici and voilà?” or ask questions in your native language – irreplaceable when you don’t know a single word in the language.
Problem: it’s not perfect! The issue lies with Whisper and is not directly caused by my product. Here’s poor Ben Tossell experiencing the same issue:
But first-time visitors don’t know nor care about this (nor should they!) So I got a lot of flak from people saying that speech recognition was junk. And, when one thing does not work, people assume that the whole thing is junk – throwing the baby with the bathwater.
Yet I could have avoided the issue entirely! I support monolingual speech recognition as well, which works much better. All I had to do was set that as the default! 🤦♂️
In short: I let a cool but not-reliable feature give a wrong first impression to probably half of users.
When stupidity takes over
I had put a lot of effort into premature optimization making sure my site could take a lot of traffic. My site runs Python on Heroku and I had made sure I’d be ready for a surge of traffic from Product Hunt, Reddit, Hacker News or whoever would feel interested enough in my humble project.
I was prepared for traffic… yet I fucked up at the last minute. Doubly so…
- I forgot to configure the dynos in Heroku to be able to handle more traffic.
Fine you say… You can just do it later if you see response times go up, right? But no:
- I had an appointment at the doc to treat allergies and, instead of postponing my post to Hacker News, I just went to the doc anyway. Result: I was AFK for most of the rush hours. n00b!
Is this self-sabotage? Maybe. ”Hey, ChatGPT! I want you to act as my shrink…”
What surprised me 😮
Unlike on Reddit where a lot of comments quickly turned into mini-discussions, on Hacker News comments were mostly one-shot. I replied to everyone but most people did not reply to my comment. I was looking forward to more back-and-forth.
I felt relieved the moment I clicked the ‘submit’ button. Overthinking breeds worries while action is liberating.
What I got from posting on Hacker News
- A record in traffic.
- A better sense of which features to prioritize.
- A boost of motivation. I’d do what I do now no matter what – because passion. But seeing people use your product is energizing! ⚡️ And beyond the peak, it’s great to see that, at any one time, there are 20 to 40 simultaneous users on the site since the post.
- Media coverage. From Poland, from India’s second most-widely read English newspaper, from a small YouTube channel covering AI by using an AI… And also from Ben’s Bites, a newsletter I’d wanted to be on for months! And it’s just so encouraging, because if a few outlets pick it up it may be only a matter of time before more do so.
- 143 signups to the free plan. So far 0 sale from it, so I’ll hold on a bit for the champagne.
- A dozen followers on Twitter.
- Contacts. A few people offered to help or discuss the product. For example:
- the CEO of a successful tech company that went through Y Combinator offered his advice, suggesting I apply to YC, saying “You've clearly got something people want (based on the comments)” and that this will help me get the product to take off;
- a top executive from a competitor backed by dozens of millions of dollars in funding offered to talk;
- a talented iOS dev offered to partner up.
And a little boost of self-esteem maybe? If all those smart people gave me a little bit of time and attention, I may just well be doing something right.
If I were to do it again…
♫ If I could turn… turn back the wheels of time… ♫
- Do your homework then let it go. Be it product development, crafting your post, or any task that requires time and attention. But once you've done your due diligence, don't overthink it. It's out of your hands how people will react. I probably could have launched a month earlier for 80% of the impact and maybe even relaunched by now.
- Use a check-list, in particular:
- Test, test, and test some more! Get out of your head and try on a new device, ask new people to try it, do hallway usability testing… See things from a new user’s eyes.
- No launch without being in-person, at the office, monitoring how things go. Duh!
Conclusion
A peak in traffic from Hacker News is nice but will neither make nor break your product. It’s just one step in the journey of being a one-man startup. So if you're a maker, go build those features, but then make sure you show them to the world to get feedback! Rinse, repeat.
A big Thanks! to everyone who posted feedback and talked about the project! Here are the stats for the first three hours:
UPDATE: actually, since writing this, things got bigger, with the site spreading through Twitter, Telegram and Instagram in Arabic, Chinese, Russian, Ukrainian and more languages. Join my adventure on Twitter as Gliglish navigates the unpredictable seas of start-up life!