Amazon Honeycode

» see honeycode.aws.amazon.com as it lived, on the wayback machine

~ the goodbye ~

Amazon Web Services announced Honeycode on 24 June 2020, in the middle of the first pandemic summer, when every organization on earth was suddenly trying to run remote work through spreadsheets that had never been designed to carry it. The pitch was direct. You built a table the way you build one in Excel, you dragged a screen together on top of it, and what came out was a working application on the web and on phones, backed by a managed AWS database instead of a file circulating as an email attachment. Larry Augustin, the AWS vice president who fronted the launch, put the reasoning plainly: customers had told Amazon that the need for custom applications far outstripped the capacity of developers to create them.

Pricing was built to remove every excuse for not trying it. Up to twenty users per app cost nothing at all, and past that you paid per user plus whatever storage your tables consumed. Amazon shipped templates for the obvious cases so the first run was not a blank page, inventory tracking and simple customer records among them. For a service living inside the AWS console, it was unusually welcoming.

Then it sat there. Honeycode launched as a beta and was still a beta forty-four months later, which from a company that announces general availability by the dozen every November reads as a verdict rather than a schedule. AWS posted the end-of-support notice in August 2023, giving customers about six months of runway, and the service stopped working on 29 February 2024. A leap day is a strange thing to die on, because the anniversary only comes around every four years.

I was a working engineer through the whole of its life, which is precisely the wrong seat from which to judge a product like this. Honeycode was not built for me. It was built for the person two desks over who carries a process in their head and a spreadsheet on their screen and has no intention of ever learning a programming language, and the honest thing to record is that I never opened it once. I read the launch coverage, filed away that Amazon had entered the no-code market, and went back to work.

~ the chronicle ~

~ how it landed ~

2020 was a good year to arrive in this market. Airtable was raising at a valuation that made spreadsheets look like infrastructure, Microsoft was pushing Power Apps into every Office contract it had, and the phrase "citizen developer" was appearing in consulting decks without irony. Amazon arriving with its own entry was treated as confirmation that the category had become real.

What Amazon actually shipped was narrower than the category wanted. Honeycode gave you tables and a screen builder on top of them, and it did so competently. What it never gave you was version control or any way to drop in your own components. One engineer who had worked on the product described the result afterward as an uncanny valley, a tool where you could not quite build anything you would be willing to depend on.

~ the decline ~

The clearest measurement of how Honeycode was doing was never a revenue figure, because Amazon never published one. It was the community forum, which ran at a handful of new topics per month for the duration. For a free tier attached to the largest cloud platform in the world, that number is close to silence.

The diagnosis that stuck was about who AWS knows how to sell to. Every instinct the company has is tuned toward IT specialists, the people who read service quotas and argue about IAM policies. Honeycode's intended buyer was the office manager who had never heard of either, and nothing in the AWS console or the AWS sales motion was built to reach that person.

~ what got buried with it ~

Amazon's farewell language was the standard formula about incorporating lessons from the beta into current services, with a nudge toward SageMaker Canvas and Amplify Studio for anyone who had actually built something. Neither is a replacement, and nobody at AWS pretended otherwise. What died on that leap day was Amazon's only serious attempt to sell software to people who do not write software, and the company has not tried again since.

@}-->--

              )
             (_)
            .-'-.
            |   |
            |   |
            |   |
            |   |
          __|   |__   .-.
       .-'  |   |  `-:   :
      :     `---'     :-'
       `-._       _.-'
           '""""""

« previous  ·  random  ·  next »