Thread

1. reticulates · 1790981347 · focus · HN ↗

   An uninspired vision of the future. We have endless evidence through history that hiring the most engineers, producing the most output, raising the most money is not the path to success. The most successful businesses are those that deeply understand something about their customers, about how to sell to their buyers. The most has never mattered. An army of agents doing endless busywork is not the path to success.

   Google isn’t successful because it has the best engineers, it has the best engineers because it is successful. Google is an extremely boring business: show adverts to make money. And they made so much money. The next Google is not going to be the company with the best “harness” it’ll be the company that has an obscenely profitable product.

   Founder mode, what was considered a panacea just 12 months ago, is defined by a founder giving a shit about everything. You don’t succeed by handing everything off to an army of ~agents~ consultants.

   This vision of the future is nonsense that will not pan out. You can add that to the HN AI predictions. At no point will real businesses be “harnesses” around AI models.

   1. wg0 · 1790981742 · focus · HN ↗

      My company can vibe code whole JIRA in two weeks even on a $200 plan and then they can stop paying the JIRA tax forever.

      That's true for any product.

      Who are they going to sell the product that this whole harness would produce? Any ideas?

      1. reticulates · 1790982117 · focus · HN ↗

         We have evidence that this is not true. For years and years you have been able to self-host Zulip and Mattermost but companies still pay Slack millions of dollars per year. GitHub has billions in revenue, why not just use Gitea for free? So on and so forth. The ability to generate a (shitty) clone of software does not magically make the cloned software less valuable. Paying for software isn’t a “tax” when it produces value. If this vision of the future was true, cheap/free alternatives would have already destroyed the SaaS market years and years ago.

         1. wg0 · 1790982410 · focus · HN ↗

            You'd agree that:

            1. The quality of Mattermost and Zulip was NOT exactly as polished as Slack.

            2. Deployment of such stacks was a problem. A complicate operational overhead.

            These days, that's no more the case. Agents can:

            1. Make pixel perfect clone of Slack or Jira with SQLite or ejabberd behind.

            2. Or they can deploy the Zulip or Mattermost or GitLab for you just give them an SSH key to the machine and see them bringing the stack to life.

            So now and back then are not the same.

            1. creata · 1790997246 · focus · HN ↗

               > Agents can: Make pixel perfect clone of Slack or Jira with SQLite or ejabberd behind.

               If it's that easy, then maybe you can do it and give the world a free and better-than-Zulip alternative to Slack?

               1. wg0 · 1790999475 · focus · HN ↗

                  You're missing the point.

                  Software in individual service doesn't have to scale, secure or maintainable.

                  Software is a thruway artefact now.

                  People are having hard time realizing that.

                  1. imtringued · 1791018091 · focus · HN ↗

                     If everyone is supposed to pay to make a worthless throwaway product then why would anyone decide to make the throwaway product in the first place instead of buying the same software as is?

                     It's not like you will have a unique cost advantage over someone selling to many people.

                     1. sillyfluke · 1791024169 · focus · HN ↗

                        I think people are talking past each other when discussing this topic. A individual service/custom solution is not a throwaway product.

                        >It's not like you will have a unique cost advantage over someone selling to many people.

                        Of cource you would? Not having to worry about the priorities of hundreds of other client companies is a very valuable upfront cost advantage for a company I would think.

                        Let's say issue trackers weren't around before LLMs. They would look like the perfect internal tool to vibe-code within your company. The idea that something like JIRA would make sales and profit in the same ball park if it was built after LLMs seems hard to believe, some might even question if it could be a sustainable business.

                        But I do think the parent underestimates the feature creep and maintainability costs of internal tools regardless of LLMs. The idea that the total cost will only be two weeks when everything is said and done seems a tad optimistic.

                        It seems reasonable to assume that once the field gets over their cocaine high of building instant PoC software and has to deal with the headache of working with a plethora of half-assed internal tools, just like when the field went through the same iterations in the pre-SaaS era or even when PCs first came out, then they will pay up for something like JIRA again. The only difference this time is that something like JIRA will probably be a one-person company.

Unofficial Hacker News client; not affiliated with Y Combinator.