KurSix

I think the lesson is that redundancy can exist at different layers

The contrast with modern software development is striking. Today we often rely on fast iteration and patching problems in production. Spacecraft software is the opposite

The XPath idea sounds great in theory, but it falls apart in a second on the modern web. Most sites (React/Vue/Tailwind) generate classes like div.flex-col.xg-9a, and the DOM structure completely changes on every single deploy. The agent will just get stuck trying to write an XPath that instantly breaks on the very next page refresh. Feeding it the visual state like the author does is way more reliable

Google is never going to upstream Chromium code that lets an external API arbitrarily freeze V8 and the render loop, purely based on the security model and stability requirements of a consumer browser. Your only real path forward is maintaining a custom patchset on top of stable releases, exactly like Brave or Electron do. Just be prepared that Claude won't save you when they inevitably rewrite the Blink architecture again

Finally someone realized that CDP just doesn't cut it for agents and dug straight into the engine. Hard freezing JS and the render loop solves 90% of the headaches with modals and dynamic DOM. Architecturally, this is probably the best thing I've seen in open source in a while. The only massive red flag is maintaining the fork - manually merging Chromium updates is an absolute meat grinder

To answer your question: this should 100% just be tooling on top of an existing language rather than a custom syntax. By ditching the mainstream, you're not just losing parsers, you're throwing away profilers, debuggers, LSP support, and millions of crates. Just write a set of macros and a strict custom linter for Rust that forces the agent to write verify blocks and declare effects. Maintaining your own custom transpiler to Rust is gonna bleed all your dev time dry long before this project ever gets off the ground

Works fine for a one-man pet project, but the second you hit a real codebase and dozens of commits a day, that nightly job is just gonna choke on the context window or slam into Anthropic's rate limits. You really need to be streaming agent state into a DB on the fly via WebSockets, not piecing it together once every 24 hours

There's no off-the-shelf "Jira for AI" product out there right now tbh. Most sane teams are either hacking together thin SQLite + WebSockets dashboards, or just sucking it up and staring at raw stdout across 5 tmux panes. If you need a self-hosted setup with tracing, Langfuse should do the trick and cover like 90% of your pain points. Honestly, if you actually polish this control plane concept and open-source it, the demand will be massive

That's exactly the trap with any LLM agent in prod. To completely drop the human-in-the-loop, you need a confidence level that current transformers simply lack architecturally. The second the model silently closes a release ticket because someone made a sarcastic joke about a rollback in the thread, you'll be the one rushing to rip that tool out of your stack

The update story has definitely been Motorola's weak spot

Yeah, this is one of the more frustrating industry-wide trends, and Motorola isn't alone in it

What's interesting about this announcement is that it implies actual cooperation instead of GrapheneOS having to reverse-engineer and work around OEM limitations. If Motorola is willing to provide proper hardware support, documentation, and not cripple things like verified boot or alternative OS installation, that's a completely different dynamic

The real question is whether Motorola is willing to accept lower short-term margins (and possible carrier friction) in exchange for long-term brand differentiation

If Motorola actually delivers unlockable, well-supported hardware that GrapheneOS endorses, this could finally break the Pixel monopoly in the hardened Android space

Yeah, logging "Hello World" from a content script takes 10 minutes. But then you run into Manifest V3 service worker lifecycles, bypassing CSP on the host page, and syncing state across tabs, and suddenly your "quick one-hour project" turns into a completely burned weekend

Cloudflare and Datadog are already training models to detect agent-like mouse movements and DOM interactions. They'll lock this down easily, and we'll just end up in an ML vs ML arms race. It's still way too easy to fingerprint a script vs a real human just by looking at input latency patterns tbh

Dumping a continuous firehose of Slack messages into an LLM context is a guaranteed way to bankrupt the company. Plus, the latency is a joke: by the time the agent parses the thread and hits the Jira API, you could've manually dragged the tickets across the board three times over yourself

Just because Claude hacked together a panel on top of Jira doesn't make it a scalable pattern. Most of these on-the-fly generated UIs are slow as hell, hijack native page events, and leak memory like crazy. It's fine for a personal PoC, but rolling this out to an entire team is pure insanity

Maybe that’s actually one of the biggest reasons researchers care so much about earlier diagnosis

Even without a cure, there's still a lot of practical value in knowing earlier

It's less like discovering the asteroid exists, and more like refining its trajectory enough to know where and when to intercept it

I get the logic, but I think the emotional side is a lot harder than the asteroid analogy makes it sound

I'd be curious to see how it performs outside of a specialist neurology setting and in larger, more diverse populations

I think it reflects a broader difference in philosophy: US sports are optimized as media products first, sporting events second, whereas in Europe it's historically been the other way around

Advertising usually isn'tt trying to create a behavior from scratch, it's trying to redirect or prioritize behavior that was already likely to happen

I think this is true if you evaluate it purely as a performance channel, but I suspect most Super Bowl buys aren't competing with search/social on the same axis

I feel the exact same way. Trying to cater to the "no-code" crowd is blurring the product's focus. It seems they've stuffed the system prompt with "be creative and explore" instructions, which kills determinism - so now we have to burn tokens just to tell it: "Don't think, just write the code"

That holds up for code generation (where tokens fly by), but not for tool use. The agent often stalls between tool calls, and those are exactly the moments I need to see what it's planning, not just stare at a blank screen

It feels like they're optimizing the UI for demo reels rather than real-world work. A clean screen is cool when everything is flying, but when things start lagging, I need verbose mode to see exactly where we're stuck and if I should even bother waiting

If they're aiming for autonomy, why have a CLI at all? Just give us a headless mode. If I'm sitting in the terminal, it means I want to control the process. Hiding logs from an operator who’s explicitly chosen to run it manually just feels weird