On 1 November 2026, a Hacker News user called roansh posted HackerBooks, a site listing about 16,500 books that at least two different HN commenters had mentioned since 2007. Each book links back to the comments that mentioned it, and the data can be downloaded. By their own account, Claude (Opus 5.5) wrote almost all of it, including the data pipeline and the site, while they “mostly steered and made calls, bought the domain name”. The post passed a thousand points.

What got people excited

The line that drew the attention was “Claude wrote almost all of it.” Someone saw Karpathy’s HackerTube, wanted the same thing for books, and ended up with a working site over 16,500 titles without writing the code. The pipeline runs on Python and DuckDB over the HN dump. The site is plain JavaScript and static JSON files on Cloudflare. The builder said the first version was clunky and needed nudges on performance, but the main instruction was “make it feel like hackertube, but for books”.

So I ask my usual question. Was there a need? Yes, a small and personal one. One reader wanted a list, built it for themselves, then shared it. What changed is the cost of getting from that want to a working thing. What did not change is everything after. The site counts mentions, so a book praised and a book trashed score the same, which the builder admits. It has no customers, no revenue model beyond affiliate links, and no promise to anyone. That is fine for a personal utility. It would not be fine for a business.

Where the shift is real

My view is that personal tools are where vibe coding is real. A normal person can now build fairly competent software for their own use. I do this myself. I have replaced many tools and scripts with ones I build with Claude. The fallout when one breaks is small, because it only touches my own data. HackerBooks is a clean example. One person, one want, one working site.

Now put that next to an operator. A shop owner who wants a sheet that flags which products are about to run out. A clinic manager who wants the week’s no-shows in one list every Monday. A marketing team that wants last month’s numbers pulled from three exports into one page. Each of these used to be a small subscription, bought from a company that built it once and sold it a thousand times. Now the owner can describe it and get something that works for their own case, on their own data, quickly.

That is why I expect consumer utility software to be hit hard. If your product is a utility that one person uses on their own data, your customer can now build their own version by steering, the way roansh did. Your competitor has become your own customer.

Where the gap stays

B2B software still has a big gap that vibe coding cannot close. The moment software touches other people’s money, orders, bookings or personal data, the hard part moves out of the code. It moves into alignment on what the thing should do, the approach, the security, and the integrations with systems the business does not control. Every build I have done needs that back-and-forth, and the AI cannot supply it without the context and experience of the person steering. You do not get it from one prompt.

Look again at the HN post. The builder steered and made calls. For a personal site, those calls are small: which books count, how the page feels, how fast it loads. For a clinic’s booking system, the calls are what happens when a deposit fails, who gets to see a patient’s phone number, and what the system may promise on the clinic’s behalf. Those decisions need someone who knows the business. They are the product.

With AeroChat, the part that takes the most work is the integrations with each market’s payment processors and logistics partners. Those are relationships and testing, and you can’t vibe code your way into them.

What to do

If you sell a simple utility, take stock honestly. List what your product does that a motivated customer could not get by describing it to an AI. If the answer is nothing, your price will fall towards zero, and you need to move towards the data, the integrations or the relationships your customers cannot build themselves.

If you run a business, build your own small tools. Keep them on your own data, where a mistake costs you little. Wherever the software touches a customer, a payment or a record you must get right, buy it, or build it with someone who has the context and will carry the back-and-forth with you.

Source: Hacker News: Show HN: HackerBooks – HackerTube, but for books mentioned on HN