No, because it's a ghost. Ghosts can't schlep. On the bright side, you don't have to feed them. Current status: putting scaffolding in place to see if I can convince Shelley to be better about...
No, because it's a ghost. Ghosts can't schlep. On the bright side, you don't have to feed them.
Current status: putting scaffolding in place to see if I can convince Shelley to be better about generating UI mockups. It sort of works.
"So far" doing a lot of heavy lifting They are banking a LOT of money in a very short time that these "teammates" are scalable 10xers who are the second coming of the indistroal revolution....
"So far" doing a lot of heavy lifting
AI teammates could present a $6 trillion global opportunity by accelerating productivity and boosting skills and creativity.
They are banking a LOT of money in a very short time that these "teammates" are scalable 10xers who are the second coming of the indistroal revolution. Someone going to pay the piper one day, and if it's not you it'll be all of us (involuntarily).
I would argue that AI is currently in the first phase of enshittification. Companies are offering amazing services at dirt‑cheap prices, losing money on customers today but betting on gaining a...
I would argue that AI is currently in the first phase of enshittification.
Companies are offering amazing services at dirt‑cheap prices, losing money on customers today but betting on gaining a majority market share tomorrow.
If this is correct, prices coming down in the future is highly unlikely. In fact, AI is probably the cheapest it will be for a long time if it follows this pattern.
Well, I'd take the other side of that bet. Companies have strong incentives to reduce their own costs and that has been the trend so far, both in general (Moore's law) and specifically for AI. The...
Well, I'd take the other side of that bet. Companies have strong incentives to reduce their own costs and that has been the trend so far, both in general (Moore's law) and specifically for AI. The whole industry is supply-constrained, so if they can increase efficiency, they can sell more. Also, there is strong competition both between American firms and with Chinese AI firms. This is unlike Bitcoin, which automatically rachets up the difficulty so all efficiency gains are lost, and more like what normally happens with industry driving costs down for new technologies.
If we're comparing like to like, today's frontier models will be outdated in a year and there will be newer models that can do about the same thing for cheaper, like happened with Google's Gemini 3 Flash.
What also happens, though, is that people find ways of increasing usage as costs go down. This has definitely been the trend in software development with coding agents. I've certainly increased my own usage significantly since discovering the joys of programming using a coding agent.
The author would have benefitted from an AI coworker giving this a brief review.
The author would have benefitted from an AI coworker giving this a brief review.
But the framing also obscures important questions about control, accuracy, and trust. A teammate who “hallucinates,” who “fails at 70% of basic tasks”—that’s not quite the collaborative partner the marketing suggests.
While many vendors are sanguine, it’s reasonable to be suspicious of this horizontal framing because for all it’s descriptive potential it also obscures important questions about control, accuracy, and trust. A teammate who “hallucinates,” who “fails at 70% of basic tasks”—that’s not quite the collaborative partner the marketing suggests.
No, AI will not bring bagels to standup, because the AI knows that standups are intended to be less than 15 minutes, and adding bagels into the mix will waste a significant amount of setup time. I...
No, AI will not bring bagels to standup, because the AI knows that standups are intended to be less than 15 minutes, and adding bagels into the mix will waste a significant amount of setup time. I asked Copilot. It also suggested making food available in some other location after the meeting. So maybe AI would order bagels, but not to be served at the meeting.
Although I'm sure some differently worded prompt could get you the opposite answer.
That's quite a good idea actually. I wish we had done it that way. The informal conversations over bagels afterwards would probably be at least as useful as the standup.
That's quite a good idea actually. I wish we had done it that way.
The informal conversations over bagels afterwards would probably be at least as useful as the standup.
No, because it's a ghost. Ghosts can't schlep. On the bright side, you don't have to feed them.
Current status: putting scaffolding in place to see if I can convince Shelley to be better about generating UI mockups. It sort of works.
Just for the sake of nitpicking:
You actually have to pay them with money and attention.
Yep, true. I find it fun, though, and the costs are reasonable so far.
"So far" doing a lot of heavy lifting
They are banking a LOT of money in a very short time that these "teammates" are scalable 10xers who are the second coming of the indistroal revolution. Someone going to pay the piper one day, and if it's not you it'll be all of us (involuntarily).
And it still won't bring bagels.
I expect that they'll find businesses to pay for fancy features, but the basic tools (like Claude Code today) will become less expensive over time.
I would argue that AI is currently in the first phase of enshittification.
Companies are offering amazing services at dirt‑cheap prices, losing money on customers today but betting on gaining a majority market share tomorrow.
If this is correct, prices coming down in the future is highly unlikely. In fact, AI is probably the cheapest it will be for a long time if it follows this pattern.
Well, I'd take the other side of that bet. Companies have strong incentives to reduce their own costs and that has been the trend so far, both in general (Moore's law) and specifically for AI. The whole industry is supply-constrained, so if they can increase efficiency, they can sell more. Also, there is strong competition both between American firms and with Chinese AI firms. This is unlike Bitcoin, which automatically rachets up the difficulty so all efficiency gains are lost, and more like what normally happens with industry driving costs down for new technologies.
If we're comparing like to like, today's frontier models will be outdated in a year and there will be newer models that can do about the same thing for cheaper, like happened with Google's Gemini 3 Flash.
What also happens, though, is that people find ways of increasing usage as costs go down. This has definitely been the trend in software development with coding agents. I've certainly increased my own usage significantly since discovering the joys of programming using a coding agent.
The author would have benefitted from an AI coworker giving this a brief review.
No, AI will not bring bagels to standup, because the AI knows that standups are intended to be less than 15 minutes, and adding bagels into the mix will waste a significant amount of setup time. I asked Copilot. It also suggested making food available in some other location after the meeting. So maybe AI would order bagels, but not to be served at the meeting.
Although I'm sure some differently worded prompt could get you the opposite answer.
That's quite a good idea actually. I wish we had done it that way.
The informal conversations over bagels afterwards would probably be at least as useful as the standup.