Post

Wrote the intro for this week's Premium but wanted to share in advance, because people need to start being realistic about what Claude Code/Codex can actually do, and how they are not replacing All Software Companies in a few prompts (and no, they will not get there either).

Ed Zitron@edzitron.com

Here is an annotated version of the Citrini Memo with my own intro. It is analyslop - scare-fiction written to ingratiate AI boosters and analysts/traders with tales of ultra-automation and socialist data center policies. Shameful that the markets reacted at all. www.dropbox.com/scl/fi/1p1n0...

26 reposts186 likes

This sounds like an LLM paradox: It saves you a lot of time by writing something basically instantly, but to be sure it's correct you would need to invest around the same amount of time it would have taken to do it yourself.

Hey Ed, I've actually talked with engineers in ATX about how Claude is being used to replace staff, and the answer isn't just "Machine do work" but that it allows for off-shoring of the work more easily. The company removed ~6 Americans in Austin for 2 engineers in Egypt that will use Claude.

The most common and realistic use case is coding assistance. You need a human who knows what should be done, and an agent helps you out. An entire vibe coded app is a nightmare, it *can* be done but abstraction, scale, and endless possibilities become overwhelming for human and agents

Asked their sales team the other day if there was a way to 100% enforce specific tests. Pointed me to making skills and rules. Went to troubleshooting in their own docs — Skill not triggering? Check the description includes keywords users would actually say. It’s all just prompts in a code casino.

Others have pointed this out, the big tell is if Claude or whatever could create a viable salesforce clone in an hour (or even a few weeks), they wouldn't be selling it. They'd keep it for themselves or setup a subsidiary to be the salesforce clone & compete w/ a fraction of the expenses (salaries).

But nobody wants a Salesforce clone for their small shop. They want 10% of what the big SaaS offers. I've done this (on a farm) where we just wrote apps to handle our basic needs. It cost way too much a month to put our 175ish employees on those. Plus the issue of logins for seasonal employees...

... sometimes the simple app that isn't fancy is the right solution. Instead of training your tiny user base you tailor the app to their needs and wants. It's super fun. And holy crap I would have CRANKED some of those with Claude Code.

I feel like I've had a version of this discussion related just to websites, as a modest dev, for at least 20 years as various friends/family came to me to build them the magic website that would make them rich, me always politely telling them "it's a bit more complicated than that" endlessly 😂

There are a lot of people legitimately emotionally excited that THEY will be able to try it themselves for the first time, because they missed the dot com days the last time through. Human excitement is a force of its own. There will be some winners, everyone wants a ticket. Nothing else to do, eh.

Indeed, everyone is convinced that the only thing stopping them from creating the next Facebook etc is they just need ~the idea~ & the rest will just fall into place. Every time I've tried to show them how even the most basic WordPress site can be set up, they are so lost. Dunning-Kruger in effect!

The most important service that SaaS companies provide to large organisations is… accountability. The actual software that the Salesforces of the world is usually pretty basic; their value is being legally, auditably reliable at scale.