Can also say the year that subscriptions go down, because now you can finally build your own systems exactly to your liking. Save money and have something that works just for you. LLMs are awesome.
good question. maintenance is cheap with tools like astra/opus. also, these aren't production level applications, so I don't need to worry about it being "perfect" (though there's no such thing)
What happens when these companies decide to start increasing the cost/token, exceeding the cost of hiring an engineer or paying for an external service in the first place?
Cost per intelligence-adjusted token has been in free-fall since GPT-3 landed five years ago [0]. I see no compelling reason to think we'll see that trend suddenly stop, much less reverse.
0. "The cost of achieving a given level of AI performance has fallen about 47% per quarter since 2023, or 13× per year."
<a href="https://epoch.ai/publications/the-plunging-price-of-thought" rel="nofollow">https://epoch.ai/publications/the-plunging-price-of-thought
Or they optimize the model for just one more hit, er, prompt. All these AI companies are backed by the same companies optimizing revenue via addiction.
So, same end result (cost exceeds hiring a new engineer), but achieved differently.
The counter, I suppose, is the managers have to deal less with engineers telling them their ideas aren't good or well thought out.
Ah yes, the ye olde "it doesn't need to be perfect" response to serious, legitimate questions.
It's always about cheaper, better, faster. And when the thing being built extends beyond a single person, and still when working alone, conflict shows up in balancing the risks between these. The Strawman of Perfection is just that.
"It doesn't need to be perfect" is incredibly dismissive of the reality of the situation. Ignoring reality seems like a poor problem solving strategy.
No one is ignoring reality. The reality is that my internal tools don't have users besides myself and as long as they do what I want them to do, I'm perfectly happy with them.
We have lived with lackluster B2B and enterprise tools for so long that we have become numb to how bad they are.
Being able to build something that works just how you want with minimal investment is a huge value add.
I ditched web analytics platforms and just send events in jsonl format to disk.
I have a regularly scheduled job that imports it all into a duckdb database and updates a very custom report. I can also ask random questions like “is there a cohort of users who use feature A and B?”.
yep, excellent use case! I did something similar with my gmail recently, because I hate using gmail's UI and I also hate that I don't get access to all my data, but I've begun to regularly backup all my email into a local postgres DB (including attachments). AI has also allowed me to get a full backup of my entire email and archive it onto my own storage. and the best part of this: I might be able to finally get off gmail
As long as the phase of the software and users are lighweight - agree these internal tools can be built, but as scale comes that is when the challenges show up is what I have observed.
If you're paying a subscription for anything, it's unlikely to be something you'd want to replace for an internal tool. Yeah, you can easily vibe code the basic functionality of a Slack/Jira/Confluence/Ashby/Salesforce replacement in a workday now, but 1) there's no chance you capture all functionality needed by your business on the first pass, you're going to be getting inundated with feature requests as soon as you roll it out, and 2) the ongoing maintenance cost is going to be much higher than expected. Not just maintaining a server and database, which is cheap, but training people how to use your tool instead of the industry standard one. There's a reason Atlassian and Salesforce stocks have recovered after the theorized "SaaSocalypse" which never materialized, it's just not an overall cost savings to build this stuff yourself.
Where internal tools shine is internal operational workflows that are being done manually. You're not replacing Slack or Salesforce, you're replacing the current process where a developer needs to SSH into the server to run a SQL query, or where the Operations team fills out a Google doc which the Finance team pastes into an Excel spreadsheet which runs a bunch of VBA calculations.
saadn92 · · focus · HN ↗
phinnaeus · · focus · HN ↗
saadn92 · · focus · HN ↗
ironman1478 · · focus · HN ↗
istjohn · · focus · HN ↗
0. "The cost of achieving a given level of AI performance has fallen about 47% per quarter since 2023, or 13× per year." <a href="https://epoch.ai/publications/the-plunging-price-of-thought" rel="nofollow">https://epoch.ai/publications/the-plunging-price-of-thought
oblio · · focus · HN ↗
oblio · · focus · HN ↗
What do you mean? They need to recover hundreds of billions in sunk costs?
spit2wind · · focus · HN ↗
So, same end result (cost exceeds hiring a new engineer), but achieved differently.
The counter, I suppose, is the managers have to deal less with engineers telling them their ideas aren't good or well thought out.
saadn92 · · focus · HN ↗
1.) open weight models gain more popularity 2.) i'm making more money from these models than paying for them
if either happens, then I'll feel good about it either way
spit2wind · · focus · HN ↗
It's always about cheaper, better, faster. And when the thing being built extends beyond a single person, and still when working alone, conflict shows up in balancing the risks between these. The Strawman of Perfection is just that.
"It doesn't need to be perfect" is incredibly dismissive of the reality of the situation. Ignoring reality seems like a poor problem solving strategy.
saadn92 · · focus · HN ↗
jmathai · · focus · HN ↗
Being able to build something that works just how you want with minimal investment is a huge value add.
I ditched web analytics platforms and just send events in jsonl format to disk.
I have a regularly scheduled job that imports it all into a duckdb database and updates a very custom report. I can also ask random questions like “is there a cohort of users who use feature A and B?”.
It’s super lightweight and incredible.
saadn92 · · focus · HN ↗
prasadvara · · focus · HN ↗
As long as the phase of the software and users are lighweight - agree these internal tools can be built, but as scale comes that is when the challenges show up is what I have observed.
mjr00 · · focus · HN ↗
Where internal tools shine is internal operational workflows that are being done manually. You're not replacing Slack or Salesforce, you're replacing the current process where a developer needs to SSH into the server to run a SQL query, or where the Operations team fills out a Google doc which the Finance team pastes into an Excel spreadsheet which runs a bunch of VBA calculations.
MisterMunchkin · · focus · HN ↗
mjr00 · · focus · HN ↗
You're aware that Slack has many more features besides chat, right?
This statement falls purely into the mistake #1 category.
prasadvara · · focus · HN ↗
This is the key distinction people need to understand, until then it just piles up and lot of companies main business is not software, something else.
sdcfgy · · focus · HN ↗
How far we have come!
saadn92 · · focus · HN ↗