677: I Accept the Battery Costhttps://atp.fm/677
-
@atpfm Its important to note that (I don't think) XCode isn't integrating with Claude Code, just Claude. They're using Claude's LLM, but not the agent-coding-loop infra that exists in Claude Code.
this makes me believe that using Claude Code will still be 'smarter', more useful, and more modern than whatever Apple builds into XCode.
Of course, the benefit will be that it'll have better GUI and integration with XCode itself.
I switch between Github Copilot and Claude Code in VS Code because I greatly prefer Copilot's IDE integration, but find Claude Code to be smarter, despite both using the same LLM models.
-
@secundus
I would be interested in references if you have them. I’ve been having a hell of a time finding numbers on this stuff.@niekvdpas As far as is known, AI companies do not publish detailed data, so there are more or less plausible guesstimates. Andy Masley (see link) has written several solid articles on the topic. Granted, he is just another person on the internet, but I find his argument reasonably convincing. This is not about specific numbers, but about a rough estimate of the order of magnitude of LLM energy consumption and how that compares to overall personal energy use
Using ChatGPT is not bad for the environment - a cheat sheet
The numbers clearly show this is a pointless distraction for the climate movement
(andymasley.substack.com)
-
@niekvdpas As far as is known, AI companies do not publish detailed data, so there are more or less plausible guesstimates. Andy Masley (see link) has written several solid articles on the topic. Granted, he is just another person on the internet, but I find his argument reasonably convincing. This is not about specific numbers, but about a rough estimate of the order of magnitude of LLM energy consumption and how that compares to overall personal energy use
Using ChatGPT is not bad for the environment - a cheat sheet
The numbers clearly show this is a pointless distraction for the climate movement
(andymasley.substack.com)
@niekvdpas To give an even vaguer answer. A fairly limited LLM can be run reasonably well on my four-year-old laptop. A much more capable model can be run on a machine, such as a Mac Studio. This does consume energy, but not very much. Especially given that typical LLM use does not imply sustained full-load operation.
Compared to these examples, optimized hardware or software, scale, better utilization, etc. will be more efficient.
-
@atpfm Its important to note that (I don't think) XCode isn't integrating with Claude Code, just Claude. They're using Claude's LLM, but not the agent-coding-loop infra that exists in Claude Code.
this makes me believe that using Claude Code will still be 'smarter', more useful, and more modern than whatever Apple builds into XCode.
@joshhunt Are you sure? I see two possible choices for Claude integration in Xcode ("Claude Agent" and "Claude"), and I'm not sure what they mean.
-
@joshhunt Are you sure? I see two possible choices for Claude integration in Xcode ("Claude Agent" and "Claude"), and I'm not sure what they mean.
@siracusa hmm no, I’m not sure at all. I’m not sure why they would have the distinction between the two, but I think it’s telling they never refer to it as Claude Code.
-
@siracusa hmm no, I’m not sure at all. I’m not sure why they would have the distinction between the two, but I think it’s telling they never refer to it as Claude Code.
@siracusa https://www.anthropic.com/news/apple-xcode-claude-agent-sdk
"Xcode 26.3 introduces a native integration with the Claude Agent SDK, the same underlying harness that powers Claude Code. Developers get the full power of Claude Code directly in Xcode"
"Developers using Claude Code can integrate with Xcode over MCP and capture visual Previews without leaving the CLI."
Agent SDK overview
Build production AI agents with Claude Code as a library
Claude API Docs (platform.claude.com)
"The Claude Code SDK has been renamed to the Claude Agent SDK"
huh!
-
677: I Accept the Battery Cost
https://atp.fm/677If you really don’t like AI, we have some bad news for you.
@atpfm Re the aftershow: this Yale working paper suggests that Elon's politics have cost Tesla more than 1 million sales: https://www.nber.org/papers/w34413
-
677: I Accept the Battery Cost
https://atp.fm/677If you really don’t like AI, we have some bad news for you.
@atpfm I heard your thoughts about Apple’s AI pin recording meetings and instantly thought of corporate IT. No way in hell are they allowing that without enterprise control and we all know how well Apple does enterprise tools. Just see all the consent messages from Zoom and Teams for recording to follow recording laws. And without the workplace use case, it seems that tech companies want and need the pin more than users do, just one more avenue to slop up data.
-
677: I Accept the Battery Cost
https://atp.fm/677If you really don’t like AI, we have some bad news for you.
@atpfm I guess you've already seen this... https://eletric-vehicles.com/waymo/waymo-exec-admits-remote-operators-in-philippines-help-guide-us-robotaxis/
-
@atpfm Waze interface while the tiles are dated, as more Applely notably in CarPlay where most will interact with it. Google interface is still very much Android. So I still find myself jumping between the three. Hopefully eventually all traffic incidents from Waze show up in Google, and the UI gets a bit more refined, then could see myself using that fully even if far less enjoyable than Apple Maps, the other benefits would probably make it worth it.
@atpfm Also went back and forth between Chrome and Safari over the years as hate filling out a long form only for it to fail in Safari and have to start all over. Disabling content blockers went things break or ordering helps a lot but often things still didn’t work. Eventually discovered turning off additional advanced privacy features in Safari settings has made the vast majority of sites work like chrome. But now losing out on privacy as can’t disable that on just select sites.
-
@atpfm Also went back and forth between Chrome and Safari over the years as hate filling out a long form only for it to fail in Safari and have to start all over. Disabling content blockers went things break or ordering helps a lot but often things still didn’t work. Eventually discovered turning off additional advanced privacy features in Safari settings has made the vast majority of sites work like chrome. But now losing out on privacy as can’t disable that on just select sites.
@atpfm and still the occasional time might need chrome if doing an advanced web app like Figma. Have noticed for long time safari was slower loading where second or so loading before anything shows unlike chrome that was instant. Finally did digging and discovered it was 1Blocker. With that off the sites load instantly and workbench test tools show Safari even faster than chrome. It makes sense having to block things before loading page would slow it down. So switched to off by default.
-
@atpfm and still the occasional time might need chrome if doing an advanced web app like Figma. Have noticed for long time safari was slower loading where second or so loading before anything shows unlike chrome that was instant. Finally did digging and discovered it was 1Blocker. With that off the sites load instantly and workbench test tools show Safari even faster than chrome. It makes sense having to block things before loading page would slow it down. So switched to off by default.
@atpfm and turn on for select ad heavy sites like new sites and working great now. With Chrome you get text 2fa autofill but most sites you don’t get native Apple Pay even though technically sites can now add support. So it’s a hassle. No mobile sync is major. Chromes adblockers are very limited. It’s ugly if prefer apples design language. Chromes web inspector less pretty but much better. And you don’t get captchas all the time and many more extensions. But still Safari here for now.
-
-
@atpfm Re the aftershow: this Yale working paper suggests that Elon's politics have cost Tesla more than 1 million sales: https://www.nber.org/papers/w34413
@benmattison @atpfm We’ve also known for at least 12 years that Elon was a fraudster when he started pushing Hyperloop to distract from trains, so this whole ”he was different ten years ago” is a bit tiresome.
-
@siracusa Claude Code like products are one of the most token-intensive applications since massive chunks of code (and essays) are being sent into the context window over and over. That makes profitability even harder - Anthropic started charging Cursor more and Cursor had to raise prices. (and both still lose money!)
@marcoshuerta @siracusa I’m glad to see someone raising this point.
It feels like everyone is having the conversation based on CURRENT pricing instead of the much higher pricing that seems necessary to break even. The point about Uber’s strategy is good, but misses important things: Uber is a viable end to end solution in a way that AI is not (even zealots acknowledge huge gaps) and Uber does not have overwhelming *ongoing* infrastructure costs for regular hardware upgrades and model updates. -
@marcoshuerta @siracusa I’m glad to see someone raising this point.
It feels like everyone is having the conversation based on CURRENT pricing instead of the much higher pricing that seems necessary to break even. The point about Uber’s strategy is good, but misses important things: Uber is a viable end to end solution in a way that AI is not (even zealots acknowledge huge gaps) and Uber does not have overwhelming *ongoing* infrastructure costs for regular hardware upgrades and model updates.@DavidAnson @siracusa Yes, a point Ed Zitron is always making is that the GPUs that they put in these data centers become obsolete and depreciate very fast.
(There is a whole side topic on why Uber wants to own a bunch of "robotaxis" that will also depreciate)
-
R relay@relay.infosec.exchange shared this topic
