camsoft2000

@steipete @Dimillian That’s pretty cool.

@Dimillian yo u need crabbox.sh - it includes a parallels adapter + checkpoints for macOS now.

I'm really hitting insane agentic engineering issues. 1. CPU bound, my M4 Max is 100% maxxed all day. 2. I/O bound, current SSD can't read/write so many parallel things fast enough. 3. Space bound, 1 TB is clearly not enough.

Am I the only person disappointed with the evolution of the M series Apple Silicon chips? When they released the M1 my MBA could run Xcode and other heavy processing and was super cool, rarely felt hot. But over the years they have been pushing more and more on the high performance cores and my M4 laptops are turning back into a dedicated heat source the very thing that plagued the Intel line. I really would have preferred the more balanced cores given laptops are portable devices. I get that we have extra power but using my M4 laptops is getting uncomfortable again.

@OrdinaryInds What wastes compute more than non-distributed systems?

What exactly do you mean by “waste compute”?

@bharath__2020 Totally agree. My snapshot tests are also E2E tests as they instrument the real MCP and CLI.

@camsoft2000 End to end test are more than useful in this world in the end for us to have confidence, but you rightly said that we need progressive testing layers for the agents to get that feedback quickly at each stages so it can expand its validation starting from Closest to all the way up

The biggest thing I’ve learn from maintaining a mostly vibe coded project is that having tests is critical without agent will throw regressions everywhere because it doesn’t have prior context with regard to historical decisions. You can try and codify some of those decisions but having integration tests is key, ideally snapshot, smoke and benchmarks too. The more your agent can self validate the better.

Let's be clear, the majority of MCP servers just expose tools, the same types of tools agents already use for things like web_search, edit, patch etc. The same rules apply, tools use context, that's not necessarily bad if that tool provides equal or greater value.

What about first-party tools? Context usage isn’t unique to MCPs, it applies to all tools, and Claude Code has historically used a lot of context internally too. Also, measuring “tokens used by MCPs” without considering the reduction in agent turns, retries, and exploratory reasoning feels misleading. A well-designed MCP can increase context usage while still massively reducing total token consumption and improving agent efficiency overall.

In the next version of Claude Code: run /usage to see a breakdown of which Skills, Agents, MCPs, and Plugins are using your tokens CLI today, coming to Desktop next

Looks like OpenAI need to hire rentokil to rid them of the goblin infestation. openai.com/index/where-th…

Codex are you okay dear?

@MainActorDev LOL. Okay.

@camsoft2000 Tried it. Decided I don't need this. And instead I created a CLI tool for it. Saving tokens.

XcodeBuildMCP has 5.6k ⭐️ on GitHub for a reason. It gives AI agents real control over iOS and macOS development: build, run, test, and have your agent interact with apps without human intervention. It’s free, open source, and transparent. No black boxes. No “trust me bro”. Just the most popular OSS agentic tool for Apple platform development. xcodebuildmcp.com

we are doin a thing labs.sentry.dev

@kr0der @rudrank Thank you! Much apreciated.

ASC CLI and XcodeBuildMCP are actually the 🐐 of building iOS apps thank you @rudrank and @camsoft2000

@kr0der @camsoft2000 Agreed

If you’ve not yet tried XcodeBuildMCP yet or unsure how to get started then I recently did a workshop which I recommend you watch. I cover installation, client setup, project setup and demonstrate how your agent can verify its own work on the simulator as well as debug runtime issues. Lastly we explain how Sentry can optionally help your agent self heal your project when issues are raised. x.com/i/broadcasts/1…

Do people tell you they actually care about their dev tools being written in Swift and not using runtimes? Not sure how that translates to a real value proposition? Also XcodeBuildMCP doesn’t expose all 70 tools, it’s also a CLI and has a skill. Sentry telemetry is opt-in when using the setup wizard and only capture runtime errors. So the only real point of difference from your reply is runtime and not being a Swift app. Also one is free and open source the other is not.

@pvncher @rileytomasek @camsoft2000 Skills over MCP. Native Swift, no Node runtime. Zero telemetry. 8 commands vs 70+ tools fighting for context. XcodeBuildMCP is solid. Different bet on architecture. Try the trial, see which feels right.

xcode is the ultimate use case for computer use. there are so many personal ios apps i haven't built because i refuse to subject myself xcode.

I haven't opened Xcode in WEEKS 👀 ive made a ton of changes to my coding workflow and added a few things that help me move 5x faster made a little video going over everything and ill cover why i still use both claude code AND cursor 😮‍💨

It's fascinating the change that is happening right now in the SWE space, every SaaS is adopting AI in some way trying to work out how to continue to stay relevant for the years ahead. The winners and losers will be determined by how these companies are willing to change and reinvent themselves. The SWE landscape in the next few years will look very different.

@SP_EnergyPeople So @SP_EnergyPeople ? Not a good look is it, won't get new customers with this level of service.

@SP_EnergyPeople Tried calling that number and there was No way to get to cancellations team even when selecting the correct option just chucks you to general CS and then it hung up on me 4 times!

Was miss-sold energy from @SP_EnergyPeople yesterday from sales people who promised me a cheaper rate, which I discover after signing up that is not cheaper at all and is more expensive than my current rate. In addition when following their own official cancellation instructions online the phone system has put me through to the wrong department even though I selected cancellations and hung up on me 4 times. This alone is a reason to avoid @ScottishPower at all costs. They make it impossible for me to enact my right to cancel in the cooling off period. Will be reporting to Energy Ombudsman.

An issue alert fired. By the time the dev saw it, Seer had already found the root cause. Two minutes later, there was a PR with the fix. That's Seer Autofix.