I cant keep track of these agent platforms (or whatever you call it.. is there even a term?) They seem to be popping out of nowhere and getting eye popping funding, that too in rounds that shouldnt be happening so soon. Am I going mad or is the bubble super frothy?
I'm actually curious about how you all see coding agents play out when every major model provider also provides a companion coding agent e.g. Claude Code, OpenAI Codex, Gemini cli ...
I wrote my thoughts[0] and felt whilst it is good to be model agnostic, the core has to be open because it is too important a tool to be left entirely in hand of others. It is also super simple to understand, OpenAI only has <5 tools for a coding agent to be functional.
Future of software as Agent and Code don't go well together in my experience. Future is bleak if this is the future.
Nothing against producing code that's AI but code reviewable easily by humans, This is ideal in fact. But having a ton of AI slop that's unreadable by humans is not maintainable.
There's of course applications like internal tools, or short lived software where this is useful no doubt.
The current SW toolchain / AI coding industry is seeing so much frothy behavior... Every toolmaker seemingly is raising $$$M at $$$Bn valuation. And they all seem to sell to largely the same pool of developers.
Reeks of self-driving startups getting funded out of every garage, about a decade ago.
Expecting to see consolidation / acquihiring / and straight-up running-outta-money in 3..2..1 years
Yep. I don’t see how this is sustainable. The people with the most capital will win. And increasingly, it looks like that’ll be the actual foundational model owners. OpenAI could presumably replicate everything this company is doing. How can they actually differentiate themselves?
I think the conclusion isn't as simple as "foundation model companies will just build the features of all downstream products" because focus and priorities play a big part.
If that were the case a simple example is much of software services we see today (and provide real tangible value) wouldn't exist as it's theoretically just updating data in a database.
ChatGPT pulse was replying "[Object object]" to thousands of users just now. You really overestimate OpenAI delivery capabilities, which is exactly what Sam Altman wants
I think they're trying to address your concern by pointing out how their Droids are different
>Every AI coding platform forces you to choose: one IDE, one LLM, one agent, one interface. Subscribing and unsubscribing from plans to try and stay at the cutting edge is now obsolete. Developers deserve a choice. Factory is giving choice back to developers. Droids work with any LLM, in any IDE, in local or remote, and in any interface. You can delegate tasks to Droids from your Terminal, your IDE, Slack, Linear, or on the web. For further customization, you can use headless mode to set up scripts or triggers to run Droids tailored to your team's workflow.
The Factory.ai Droid does rank well in the terminal-bench leaderboard (currently 3 of the top 5)
Whenever a new frontier model is released, 99% off the demand shifts immediately to that model. The price of the most demanded model has remained roughly constant even as prices for legacy models decay.
An ice cream shop just opened a few blocks from my house.
For weeks, they've had a line of dozens even at 9pm when I drive by.
I don't think that "everyone likes trying new things when they first become available" is unique to AI. Even ice cream stores have this effect, so I'm not sure if you just discount everything new or aren't familiar with this general phenomenon.
In terms of what model I use it’s going to be latest and greatest out of whatever my org gives me enterprise access to.
I’d probably try the factory ‘droid’ first, but if it fails to solve a problem and swapping to another one means it solves it, I’ll probably never use the factory one again.
I’d consider this pretty normal as we’re moving closer to ‘this is actually a mature stack that devs use for work’. Very few people in my company actually take interest in their development tooling and use whatever came pre-configured. (our stack is not sexy)
(Let me know if I’m incorrectly conflating the ‘droid’ concept with foundation models)
Weird how the old ice cream shops still exist, even when the market magically produces hustle-and-bustle for the newest shops temporarily when they open
Almost like demand and markets can change/grow in response to what is offered
Human nature is to buzz around anything new more than stuff we already know, and of course that dynamic will surround new and exciting things
The $20/mo plan is not the goal, that's not even close to sustainable, that's basically just a charge to verify someone is human. The goal is you reduce that 25m to 20m and you charge 1/5th of a SWE salary, let's says it's 100k and you replace 5m with AI agents @ 20% that's 100 billion, and that's just one industry. It's also one of the more difficult to replace because software engineers are actually much closer to artist than engineers.
It's the actual engineers, lawyers, marketers, medical professionals and accountants that are the real targets.
There is a lot of focus on AI assisting software development right now because they are the people most willing to use it. AI as it is now will probably create more SWE positions overall. What you will have is software designers being added by AI and it's domain knowledge to build applications that replace the other professional roles.
The cost to utilize AI is going to be much much more expensive as it is heavily subsidized right now.
Thank you, nice insight. I am struggling to envision how there will be more swe positions. Perhaps a concentrated class of software designers that nail this new skill set.
As someone who is deeply involved in this space, I fail to understand what the product does. This is genuine feedback. I signed up but not sure what I am looking at.
I work at Factory, but this is just my own thinking, not the official company line.
It’s a system for interacting with LLMs in multiple ways to do all the neat agenty things you might want to do with them. Find it convenient to work in a web browser with code actions running in a cloud VM? Sure. Prefer to work with a TUI client in your terminal (or from within Emacs, as I was using it today)? You bet. As a headless command line tool that you can use to script up cron jobs or wire into a Slack bot or run in a GitHub action? You can do that. And in each of those situations, you can choose the model you want to use, without separately creating an account with each provider.
That’s the gist of it from my perspective as an engineer. There’s more to it than that, but these are the bits that I personally love.
Maybe I'm missing something but it's not clear to me why I should use Factory's Droids instead of Claude Code -- I'm not really interested in switching models all the time.
I cant keep track of these agent platforms (or whatever you call it.. is there even a term?) They seem to be popping out of nowhere and getting eye popping funding, that too in rounds that shouldnt be happening so soon. Am I going mad or is the bubble super frothy?
There's genuine rapid innovation in this space right now.
> We are a tiny but prolific team in San Francisco (with a median typing speed of 120 wpm)
I couldn't resist.
I'm actually curious about how you all see coding agents play out when every major model provider also provides a companion coding agent e.g. Claude Code, OpenAI Codex, Gemini cli ...
I wrote my thoughts[0] and felt whilst it is good to be model agnostic, the core has to be open because it is too important a tool to be left entirely in hand of others. It is also super simple to understand, OpenAI only has <5 tools for a coding agent to be functional.
[0]: https://blog.toolkami.com/openai-codex-tools/
How much of that $50M do you think went to license the term "droid". https://en.wikipedia.org/wiki/Droid_(Star_Wars)#Trademark
Future of software as Agent and Code don't go well together in my experience. Future is bleak if this is the future.
Nothing against producing code that's AI but code reviewable easily by humans, This is ideal in fact. But having a ton of AI slop that's unreadable by humans is not maintainable. There's of course applications like internal tools, or short lived software where this is useful no doubt.
The current SW toolchain / AI coding industry is seeing so much frothy behavior... Every toolmaker seemingly is raising $$$M at $$$Bn valuation. And they all seem to sell to largely the same pool of developers.
Reeks of self-driving startups getting funded out of every garage, about a decade ago.
Expecting to see consolidation / acquihiring / and straight-up running-outta-money in 3..2..1 years
Yep. I don’t see how this is sustainable. The people with the most capital will win. And increasingly, it looks like that’ll be the actual foundational model owners. OpenAI could presumably replicate everything this company is doing. How can they actually differentiate themselves?
I think the conclusion isn't as simple as "foundation model companies will just build the features of all downstream products" because focus and priorities play a big part.
If that were the case a simple example is much of software services we see today (and provide real tangible value) wouldn't exist as it's theoretically just updating data in a database.
ChatGPT pulse was replying "[Object object]" to thousands of users just now. You really overestimate OpenAI delivery capabilities, which is exactly what Sam Altman wants
I think they're trying to address your concern by pointing out how their Droids are different
>Every AI coding platform forces you to choose: one IDE, one LLM, one agent, one interface. Subscribing and unsubscribing from plans to try and stay at the cutting edge is now obsolete. Developers deserve a choice. Factory is giving choice back to developers. Droids work with any LLM, in any IDE, in local or remote, and in any interface. You can delegate tasks to Droids from your Terminal, your IDE, Slack, Linear, or on the web. For further customization, you can use headless mode to set up scripts or triggers to run Droids tailored to your team's workflow.
The Factory.ai Droid does rank well in the terminal-bench leaderboard (currently 3 of the top 5)
Except that's not how humans work. No one is itching to try Gemini 2.5 Flash when Opus 4 is on the table.
They will go to wherever they can get the best model, driving whoever is crazy enough to offer for them for 20/mo almost immediately out of business.
There's no flat pricing model that can sustain what developers try to throw at it.
> Except that's not how humans work. No one is itching to try Gemini 2.5 Flash when Opus 4 is on the table.
You're welcome to speak for yourself and your use cases, but speaking for all humankind seems a bit much.
Whenever a new frontier model is released, 99% off the demand shifts immediately to that model. The price of the most demanded model has remained roughly constant even as prices for legacy models decay.
https://ethanding.substack.com/p/ai-subscriptions-get-short-...
An ice cream shop just opened a few blocks from my house.
For weeks, they've had a line of dozens even at 9pm when I drive by.
I don't think that "everyone likes trying new things when they first become available" is unique to AI. Even ice cream stores have this effect, so I'm not sure if you just discount everything new or aren't familiar with this general phenomenon.
In terms of what model I use it’s going to be latest and greatest out of whatever my org gives me enterprise access to.
I’d probably try the factory ‘droid’ first, but if it fails to solve a problem and swapping to another one means it solves it, I’ll probably never use the factory one again.
I’d consider this pretty normal as we’re moving closer to ‘this is actually a mature stack that devs use for work’. Very few people in my company actually take interest in their development tooling and use whatever came pre-configured. (our stack is not sexy)
(Let me know if I’m incorrectly conflating the ‘droid’ concept with foundation models)
Right, and new ice cream shops keep opening up all the time, and people are switching to the new one almost as soon as it becomes available.
Weird how the old ice cream shops still exist, even when the market magically produces hustle-and-bustle for the newest shops temporarily when they open
Almost like demand and markets can change/grow in response to what is offered
Human nature is to buzz around anything new more than stuff we already know, and of course that dynamic will surround new and exciting things
Like cutting edge AI
Consider that there’s only around 25mn people employed as anything closely described as SWE, these valuations are just baffling.
The TAM, if you got absolutely every one of these people onto a $20/mo plan is $500,000,000/mo. For reference, that’s less than Airbnb’s revenue.
The $20/mo plan is not the goal, that's not even close to sustainable, that's basically just a charge to verify someone is human. The goal is you reduce that 25m to 20m and you charge 1/5th of a SWE salary, let's says it's 100k and you replace 5m with AI agents @ 20% that's 100 billion, and that's just one industry. It's also one of the more difficult to replace because software engineers are actually much closer to artist than engineers.
It's the actual engineers, lawyers, marketers, medical professionals and accountants that are the real targets.
There is a lot of focus on AI assisting software development right now because they are the people most willing to use it. AI as it is now will probably create more SWE positions overall. What you will have is software designers being added by AI and it's domain knowledge to build applications that replace the other professional roles.
The cost to utilize AI is going to be much much more expensive as it is heavily subsidized right now.
Thank you, nice insight. I am struggling to envision how there will be more swe positions. Perhaps a concentrated class of software designers that nail this new skill set.
AI isn't actually that good at replacing software engineers, but it's going to be great at helping software engineers replace other professionals.
Wouldn’t the TAM be defined by number of agents? In the end there will be more agents coding than people.
I can wait to see what the next bublle HN wants to sell me!
LISP machines. Trust me.
The next hype to the level of AI will be quantum computing
What's the problem with that? Seems like a good thing to have people spraying money at engineers to try to find something that works.
I had this thought as well…
If there’s a firehose of money, maybe I’m the stupid one for not standing in front of it.
After all, a bubble is just a bull market you don’t have a position in :)
> After all, a bubble is just a bull market you don’t have a position in :)
That's a good one. I think I might steal it :)
The same people that are laying off engineers for the last 3 years?
Regardless I love their website's design and feel.
So aside from the integrations which do seem nice is their moat a good prompt or an I misunderstanding?
As someone who is deeply involved in this space, I fail to understand what the product does. This is genuine feedback. I signed up but not sure what I am looking at.
I work at Factory, but this is just my own thinking, not the official company line.
It’s a system for interacting with LLMs in multiple ways to do all the neat agenty things you might want to do with them. Find it convenient to work in a web browser with code actions running in a cloud VM? Sure. Prefer to work with a TUI client in your terminal (or from within Emacs, as I was using it today)? You bet. As a headless command line tool that you can use to script up cron jobs or wire into a Slack bot or run in a GitHub action? You can do that. And in each of those situations, you can choose the model you want to use, without separately creating an account with each provider.
That’s the gist of it from my perspective as an engineer. There’s more to it than that, but these are the bits that I personally love.
What problem does it solve?
"It’s a system for interacting with LLMs in multiple ways to do all the neat agenty things you might want to do with them"
This answer has way less meaning to the readers than you think
congrats to factory on the amazing product and release!
Maybe I'm missing something but it's not clear to me why I should use Factory's Droids instead of Claude Code -- I'm not really interested in switching models all the time.
Might just burst soon... https://www.youtube.com/watch?v=37aUuoRyMhM
voiceover on that video sounded like it was ai generated lool
A tale as old as time, startup CEO's and self-indulgent announcement videos with no adherence to norms of storytelling technique or pacing.