Shutting Down My Baby | Hacker News Reader

Shutting Down My Baby

fabiandietrich.com

> Like so many startup founders, I was driven by the motivation to build a product that I would like to use myself (or so I thought). My goal was to allow authors to automatically generate new creative text based on text they had written so far (basically a literary autocomplete).

Lesson learned I hope: It seems to me that the author wanted to be an entrepreneur foremost instead of solving an acutal problem. What the author wanted was their own startup with the accompanied startup experience, and so the "problem" was rather secondary and maybe even thought of after that desire?

I mean:

> It turned out that writers actually like to write themselves, and the current level of technology was just not good enough to make them really satisfied with whatever output they got from my product.

Really? It continues:

> Even though nearly everybody on this planet has heard the “machine learning” buzzword by now, the less technologically experienced hobby writers that I talked to didn’t have a real understanding of what they could expect from my product. This fundamental knowledge divide really made me think about how to reach this group of people.

Again, this really gives me the impression that you're not trying to solve a problem. You're "thinking about how to reach this group of people" to sell them your dream of having a company, not something they want or need.

I mean, to a certain extent, running a startup often means convincing people of a problem they didn't know they had. I think a more accurate simplification of their post mortem is something like:

"I had a problem and solved it for myself. I thought it was a problem others had, but it turned out most people didn't think it was a problem, and I couldn't convince them otherwise."

I was wondering the same.

It feels like a mature version of this product would be ideal for publishers and other customers of authors.

However I suspect we are not there yet technological wise and therefore he was unable/didn't attempt to sell the product to these companies.

It would be ideal for someone, but either he can't summarize his product well or it gives me the chills.. Autogenerated texts with correct form and an appearance of putting in the thought is the kind of thing that turns people off to whole industries and gets them to go back to reading things with an earlier publication date than the technology.

This seems like a product fit for content farms/SEO optimised sites that have no intention of anybody reading the output.

> publishers and other customers of authors

To do what? Auto-generate new books about a particular series when the writer doesn’t want to write any more?

Possibly with some prompting help from a ghostwriter?

It IS basically Github's CoPilot, but for plain text, no?

>"thinking about how to reach this group of people" to sell them your dream of having a company

This is actually a very effective approach if you have something to offer they both want & need.

I think the criticism was that the author of the article was more interested in "being an entrepreneur" than in offering something they want and need. In this case in particular, offering people a way to automate away things they do for fun seems to be basically the opposite of what they want and need.

I agree, but I wouldn't want to throw the baby out with the bathwater.

> The feedback I received was generally favorable, and people thought of my product as a cool gadget. But potentially paying for it? Nah. It turned out that writers actually like to write themselves, and the current level of technology was just not good enough to make them really satisfied with whatever output they got from my product.

Exactly. It's like inventing a knitting machine and reaching out to hand-knitters. You're offering to destroy through automation the thing they enjoy. The right market to reach out to would have been consumers of writing.

To the extent that this already exists, it's used to create blogspam for SEO purposes. AI-generated text is going to make the next generation of the internet so much worse.

> It turned out that writers actually like to write themselves

After reading this line, I literally went thank goodness it flopped.

I don't want to read autogenerated content no matter how good it may be.

Yes, AI can be useful in a lot of domains, but autogenerated content to read? No thank you. I like to experience what the writers themselves come up with.

And what if the writers change their style?

> AI-generated text is going to make the next generation of the internet so much worse.

This kind of thing combined with global warming would make for an interesting future.

> This kind of thing combined with global warming would make for an interesting future

AI can put out hundreds of thousands of articles telling you why global warming is a good thing. After all, nothing has meaning to an AI: not its own survival, and certainly not yours.

There's a Dilbert cartoon (which alas I can no longer find) where the punchline is "experience doesn't work like that".

In other words you can predict the outcome of someone else's endeavour, because of your own experience, you could tell them your prediction, but they'd go ahead anyway because experience has to be lived, and is seldom passed from one person to another.

This startup post-mortem seems to happen almost weekly, and despite all the literature (formal and informal) it seems like the same mistakes are made over and over again.

To the author I say, well done, you are gaining some valuable experience, and learning some valuable lessons. By all means get a job to replenish empty coffers, so that when you do another startup you apply those lessons learnt (and if that fails you'll have more experiences to share.)

For my own journey I got lucky. We bootstrapped the business with no investment, writing custom software, until we found our product niche and pivoted into selling that. I have no idea if that strategy would work today - and alas the experience of our mistakes (of which there were many) is unlikely to be transferable.

Good luck! (seriously, a big of good luck really helps a lot...)

>>> In other words you can predict the outcome of someone else's endeavour, because of your own experience, you could tell them your prediction, but they'd go ahead anyway because experience has to be lived, and is seldom passed from one person to another.

This is true in general. People still go ahead because their thinking is your situation is different than mine, and also that this time it will be different since its not you but me who will implement it.

Or, as monty python might say;

Yes, it's impossible, but imagine the possibilities if he succeeds... [1]

[1] from the sheep who thinks he's a bird, and is teaching the other sheep how to fly..

Well, because it is. No two situations are alike. If we only keep to not-doing things we've been warned about on the internet, then we'll become paralyzed and not try anything.

https://dilbert.com/strip/2007-10-14

Thanks!!!

Microsoft has incorporated this "feature" into Word and Outlook, and probably everywhere else, and it is absolutely infuriatingly stupid. Just another bit of rubbish to disable. For OP, if he had thought of it earlier he might have been able to sell it to those rich idiots. Who knows.

I suppose in rare (jargon- or abstruse knowledge-heavy writing, and activated only by ctrl+space, etc) it might be useful.

> My goal was to allow authors to automatically generate new creative text based on text they had written so far (basically a literary autocomplete). (...) The feedback I received was generally favorable, and people thought of my product as a cool gadget. But potentially paying for it? Nah. It turned out that writers actually like to write themselves

Here's what I would pay for: something that would generate a structure / a detailed lists of scenes, given a list of characters and a theme, or the beginning of the text, or maybe from scratch.

I'm an amateur writer, self-published my first novel last year, which is doing reasonably well. I like to write scenes, dialogue, interactions, and given a scene context, can sometimes write fast.

But I have a hard time coming up with what to write, what scene to write, what comes next. I generally know where the whole story is going and how it's going to end, but what needs to be written now, I don't know.

I suspect many authors are in the same situation.

Yeah this seems like it could be a good product for combating writer's block. I don't know if OP tried that angle, but from the tiny amounts of writing I've done, often I could get unblocked if I could only get a couple more sentences written - usually at a transitional point in the story.

A singer-songwriter asked me to make an app like yours.

I asked him: an app for rhyme suggestions? He said yes, but that it was more for suggesting ideas.

Rhyming dictionaries have existed for a very long time: at least back to 1775[1].

Not that an app wouldn't be useful to someone, but if they need rhymes today, bookshops (and websites) have an MVP already.

[1]: https://en.wikipedia.org/wiki/Walker%27s_Rhyming_Dictionary

He meant something like "Rhymes would be cool, but that's not what I'm telling you. I want to see ideas related to what I'm writing about."

1775 is not "a very long time" ago by most definitions.

Rhyme dictionaries are significantly older than that. https://en.wikipedia.org/wiki/Qieyun

(Ironically, old Chinese rhyme dictionaries have considerable scholarly importance because they allow us to determine whether old Chinese text probably rhymed, a task the text itself does not help with. They are less valuable for the other direction of retrieving a list of rhymes for a particular character.)

I do not know - this can be a great product for “SEO consultants”. I.e., creating blog spam.

> build a product that I would like to use myself

Yup, that's a first mistake, no? Shouldn't one build projects for others? It's a fine distinction, but getting that first customer is so important (or so I've heard).

It can work if coincidentally it's something others want and will pay for. That's sometimes the case, and those for whom it is often pontificate upon that coincidence.

On the other hand though, if you build a product that you'd like to use yourself then you might just discover an untapped niche.

If it turns out not to be an untapped niche though, or not a good one... ugh. ;)

It's actually the opposite. You always build for yourself or someone comes to you and you build a product for him. Because that shows a real need.

That's also how it worked for my product I'm very successful with (I built it for myself).

If you try to think what other people would use and build a product around that you will most likely fail.

>> build a product that I would like to use myself

> Yup, that's a first mistake, no?

No. The anti-pattern here is

"Build something you think other people would want because you would like it."

transforming into ->

"Build something you think other people would want because you would like them to like it."

(there's a subtle change)

In three or four startups I've been associated with, this was going on. In psychology it's 'projection', and I think many young entrepreneurs fall into it.

Indeed it's a common problem in digital technologies - the development of a feature by one group of people, who've become convinced another group want it, so push it against all evidence and market behaviour.

Innovators can get enchanted by their own idea, and really want to be "creatively validated" by what they see as an "audience". This has gotten worse as software startups took on the characteristics of a "new pop music" business.

Or, (with existing semi-succesful companies) they get mixed up in unconscious ulterior motives for wanting people to adopt an idea (such as displacing another product/paradigm they dislike, or effecting some kind of control over users (DRM etc).

I think the mistake is actually “like to use”, it should be “want to pay for”. People wanting to use the product is not enough, people have to need the product so much they will pay money for it. The question the founder needs to answer is, how much would he pay to have this product. How much will they pay, and why will they do that?

Reply on news.ycombinator.com