Poll: What book should I write? | Hacker News Reader

Poll: What book should I write?

56 pointsmjburgess39 comments

I keep some personal notes for areas that I work in, which could be rewritten into books -- in a tech self-publishing style.

Which book idea would you be most interested in reading? (You can vote for multiple... and suggest related titles)

View on HN

I'm the author of four nonfiction books, none about programming and none self-published, but I think some of my experience still holds.

You're doing market research. That's great, and it's a step that a lot of people neglect to do before embarking on a book-length project. A traditional publisher would expect you to do a decent amount of this in your proposal, and the thing they're looking for is, "How is this different enough from everything that's already out there that it fills an unmet need, but similar enough to books that have sold well that we can expect it to sell well too?"

The other thing that a lot of would-be authors underestimate the importance of is their platform. Anybody can come up with a great book idea, far fewer can actually write it well... but the real differentiator is how big of an existing audience you have and how well you can sell it to them.

Finally, you've thrown out a grab bag of possible titles, but which of those are you truly passionate about? If you do commit to writing a book, it's like a marriage. It's going to be misery if you aren't genuinely interested in your partner.

I have these issues in mind. The way i'm intending to solve them is via the self-publishing model. I have written large works before (including non-public books on Scala, R and effectively a couple textbooks on data science).

What I'm aiming to do here is cautiously acquire a minor second income stream by taking a lot of my already well-formed thoughts on various areas, and create small and cheap PDFs that a few people would be interested in reading.

I am very wary about "When Machine Learning Fails" as that feels like one of those magnum-opus books that takes years to write and a mental breakdown to finish. If I were to do that, I'd probably release a first chapter and mid-chapter as drafts to gauge interest.

The first chapter would be a lighlty mathematical theory of learning & philosophy discussion, which should polarise enough people into like/dislike -- and a mid-section chapter would be a semi-math. semi-prog. case study popular write up of, probably, an academic paper.

I think also the machine learning one is quite a bit riskier than the others due to how quickly the field is moving. Big failures you write about may be easily worked around by the time you finish, and even if everything is current when you finish, the book may have a pretty short shelf life.

What do you mean by "non-public books?" Are these books that you haven't released or published but have already written?

Tangent: I really thought Bruce Eckel's strategy for writing would become the norm. Especially for technical topics. Especially with publishers like O'Reilly.

(You may recall Eckel's Thinking in Java and Thinking in C++.) Eckel wrote in public, released early drafts, actively sought and incorporated feedback and errata. As a former author of shareware, this strategy resonated with me.

As an author yourself, any thoughts on why Eckel's strategy didn't catch on? eg Too much effort, publishers wouldn't buy-in, just a weird idea...?

FWIW, I consumed those drafts asap, and provided (very) modest feedback, so I was heavily emotionally invested in Eckel's success. Once available in paper, I bought many copies; at least 5 alone as gifts.

I am another self-published author. I am of the belief that the best poll you can conduct is to announce that something exists and then try to get people to invest something—time, attention, money—in it.

For example, you can announce you're writing a certain book and ask people to sign up for a mailing list to be informed about updates. Or follow the LeanPub model and offer the unfinished book for sale at a steep discount, with early adopters/patrons getting all updates.

I count amongst my biggest successes a small list of books I DIDN'T write because although people told me that they were interested in the books, mocking up a fake cover and attempting to sign people up for emails told me that the market wasn't actually receptive.

The key is that asking people to make even a small investment gives you a far truer signal than any poll.

A decent book that elaborates on this idea and gives some examples of how to pre-test is called ‘The Right It’ - https://www.albertosavoia.com/therightit.html

To be honest, some of these ideas have a fairly saturated market. If you write another python tricks & tips book or another data science starter examples book, it might sell some copies and make some money. But it probably won't contribute a lot unless you have some extremely good and new ideas and can execute them consistently.

Medium-complexity books are much more interesting but also harder to write.

"in pictures" books are also great; I feel we don't have enough of them. But they are also not very simple, because if you can't use animation, some phenomena are kind of hard to convey. If you can pull that off for graph theory or statistics, awesome!

Perhaps you should take one hard example from each domain and think about how you would write it.

I.e. How would you explain a gibbs sampler in pictures?

I have lots of iPad notes already "in pictures" for these topics, and have many versions of them as I've evolved the diagrams and examples I use... hence the idea (ie., minimising work for myself).

I'll see if a draft of my-style sells a few copies, on the topics I have to hand, and then look at outlining a larger series.

- Thanks!

That's a good idea. I would also try and show your notes/diagrams to people (both those who don't know the subject yet and those who do) to get their feedback.

That sounds like it could also be a series of blog posts, twitter posts or whatever that is then sold in book form. Depending on the popularity this might work as well - there are many examples.

Don't write anything if you don't know where your heart lies. The world could do without endless seas of copy pasta making it to the bookstores.

I feel this greatly as an author, but I also acknowledge that other people have different motivations. It's like saying "Don't write software if you aren't passionate about software."

If we're lucky enough to have worked our way up Maslow's Hierarchy to the point where self-actualization is our prime motivation, that's great! But others may want to put food on the table, or enjoy a minor amount of prestige as an author, &c.

Don't worry in this case, I have zero intention of rewriting public documentation and calling it a book.

Everything here I've spend weeks with others explaining, and so on, so the words will be my own -- and, through practice, I think well aimed.

> I have zero intention of rewriting public documentation and calling it a book

Put like that, it sounds terrible. But if I may way metaphorically... Consider a Museum or Art Gallery. There is going to be a public, free guide to the galleries listing everything in it. This is like documentation.

But there are also often human guides that can take you on a curated tour. perhaps for International Woman's Day they can lead you from exhibit to exhibit talking about the status of women during the time period represented, or focus on woman artists, &c.

Those highly opinionated, curated tours add a lot of value even if everything they show you is already listed in the official "documentation."

I think there will always be room for curated tours.

If you're going to write "When Machine Learning Fails", then the following spreadsheet might be helpful: https://docs.google.com/spreadsheets/u/1/d/e/2PACX-1vRPiprOa...

Thanks -- it feels like the most intensive one to write, but in many ways, its been my main topic of thought for several years

"When Machine Learning Fails" and commit to using a subtitle generated with a (flawed) piece of code from the book.

"When Programming Language Examples Fail" (with pictures!)

To "When Machine Learning Fails" add a chapter or two on "Sabotaging Machine Learning Models".

I'd say the second chapter will be something like "the implications of adversarial examples on the theory of learning"

In other-words, making the argument that adversarial examples show it is the properties of the datasets on which algorithms operate rather than the properties of those algs which cause ML to succeed/fail

This is the kind of stuff I had in mind when making the suggestion about sabotage:

https://arxiv.org/abs/2204.06974

When Machine Learning Fails. I'm especially interested when reinforcement learning doesn't work, what do I do then? How do I debug?

My first programming book was one that introduced several different languages at once. I forget what it was called but I remember TCL, Perl, PHP, C, maybe Bash.

It was successful in igniting the spark inside me that lead to more books and more learning.

Often, and especially these days, all people need to get started is an introduction to the environment setup, and where the docs are.

If you don't already know the answer to which book you want to write then the question you're asking is "which book do you think will sell".

The answer to that question is "who cares". No one wants to read yet another tech book written by someone who is not both qualified in and passionate about the subject.

If people don´t know what to write about,people should ask themself why they don´t read? Find a problem, fix it.

I have a question.

When you say that you are going to write a book titled "When Machine Learning Fails", is your book going to based on the boundary, the limits of machine learning or is it going to talk about obvious limitations and examples? Will it have interesting, insightful case studies, or the regular toxic, baseless anti-ML stuff?

          ° ° ° | °  °   °
     ° °     °  | °   °
      °    °    |   °
        °       |      °
       °   °    |   °  ^
       °        |      |B°
                ^
                | A
     ML Applies      ML Fails

Are we going to get stuff about points in A or trivial stuff like B?

Will this be a lesson for MBAs so they can steer away from use of ML in projects, or the limitations of ML in an ML 101 course?

Will this be on ML theory and practice deep enough to teach ML/DL practitioners new lessons and insights?

Caveat: At the time of adding this comment there are 2 points for "12 Programming Languages" and 3 for "Data Science Starter", so they are the preferred options at present.

For the 12 languages if you are looking to create some revenue from this it could be seen as direct competition to "7 languages in 7 weeks" (https://pragprog.com/titles/btlang/seven-languages-in-seven-...). So you may need to expand on what your book would provide that extends / adds to the work done in this previous book (other than comparing 12 languages instead of 7).

Since you seem to have a broad knowledge of programming languages a good book on 'how to choose a programming languge' (i.e. to which requirements fit a language (including ecosystem)). Though 'statistics in pictures' sounds nice too.

Critically evaluating programming languages sounds like a "gap in the market" -- I suspect there's a small but high-interest market in trying to put programming languages in a broader theoretical and practical context.

"The Critical Programmer: How to Choose and Use the right Programming Language"

> ext.

> "The Critical Programmer: How to Choose and Use the right Programming Language

Just use Javascript TLDR; jk

Whatever you write, see if you can imbue it with a sense of personality. I think readers are more open to this now, and it's a great way to stand out in a world of often dull, derivative books. I wish more of the technical books I read also shared the author's opinions, experiences and gut feelings (even if just in sidebars/breakouts) as well as the facts. The real world experiences of the author could add a lot of depth to a technical book.

I'm not sure if these are "book shaped" ideas, or perhaps more blog/webpage/Instagram shaped.

"When Machine Learning Fails" reminds me a lot to Matt Parker's "Humble Pi"[1]to me

[1]:https://www.amazon.com/Humble-Pi-Comedy-Maths-Errors/dp/0241...

“When Machine Learning Fails: in Pictures” could be a bestseller, with the right pictures.

[Serious] I prefer video. Not gonna sit down and read. Statistics in pictures sounds like a good idea for an animated video series. Not sure how to pay you for it though.

If you go for any of the "in Pictures" ones, I'll gladly help. I've been working out on a similar thing.

Write a book about how to choose a topic for a book.

Suggestion: Python + ML

Reply on news.ycombinator.com