Drew Ewing

In the midst of a war against ourselves from the future and nobody seems to notice shit.

@boyuan__zheng Model desperately needs steering to ritualize researching and understanding the scope of a user request, confirming it, and then executubg until all of the obvious steps towards realizing it are taken, with an adversarial step taken at each gate to keep things honest and accurate

Excited to see people try Grok Build for web dev. Our team has put a lot of effort into improving its aesthetics, functionality, and more exciting features to be expected with recursive self-improvement loop. It’s still early beta, and feedback is very welcome. Please try it out and let us know where we can improve.

@kunchenguid x.com/vitromancer/st…

i'm strongly against model companies focusing too much on harness, but i would love to hear if anyone has a strong argument for it my reason against it: if openai didn't build GPT 5.5, no one else can. this is their core competence if openai didn't build codex cli and app, we have opencode and t3code. building harness is NOT their core competence this is not saying products like claude code, codex aren't good - i genuinely think these are top tier products built by really talented people my point is - the world might be a better place if model companies focus more on their core capability and give us better, faster, safer and cheaper models, rather than competing with the ecosystem in the application layer what do you think?

@gqueiroz1313 @seconds_0 @JustJake @_mateic @theo You let Railway be your single point of failure and you are less composed in this conversation so you would be wise to point your frustration and criticisms at yourself, you failed in exactly the same way but more. You should hope your customers are more reasonable than you.

@seconds_0 @JustJake @_mateic @theo I couldn't care less if Railway wants to lose me. My priority is my own clients, who suffered a disruption until I manually migrated to my backup cloud. Expecting the service level I literally pay for isn’t being a bad customer, it’s called running a business.

It appears Google Cloud has blocked our account, and so some services are unavailable We’ve escalated this to Google and will keep people posted. Deepest apologies.

@viemccoy spud me pls

@elijahmcom @wesbos I think they built it internally for SpaceX and it kinda fucks so they are publishing. If Cursor plays out, great, they take the best of each CLI and make a better one. If not, they keep going. Seems wise.

@wesbos You think this is a hedge?

Grok's claude competitor is coming Grok build is only available to $300/mo SuperHeavy users at the moment. Anyone tried it? I've been very impressed at the speed/cost/quality of the latest xai models - specifically the speech and image ones.

Why turn lead into gold when you can turn sand into thought

@charliermarsh share the wealth

What would you do with unlimited tokens

@kilocode yea, it's cracked when it feels like working. devs are doing a great job at pushing it forward.

Grok Build 0.1 might be one of the most underestimated AI models right now. We tested it in Kilo Code by asking it to build 5 websites from scratch. Here are the results:

Qwen3.5 397b a17b is so damn good. Runs great on MLX on M3 Ultra. 4b quant is better than 5.3-codex, no bullshit. No idea what the benchmarks say, this is just real world truth.

Grok build has been avoiding long running tasks today but an hour ago decided to lock tf in… first time I’ve seen it tackle a real problem aggressively. Definitely encourage sub agent usage for complex tasks. Super effective as an orchestrator.

@skcd42 Big improvement!

Bug fixes shipping for Grok Build (release notes will be available in the TUI itself) - Copy only path on ReadTool block - Redirect stderr to /dev/null to fix incorrect TUI text interleaving - Distinguish system/user initiated cancellations - Guard against crash conditions in event loop - Prevent mouse event leaks - Interactive read file viewer - Improve pending state UI - Lower MCP use_tool output limit to 20KB - Fix for agent waiting on bg task and subagent completion - Embed read_file (images and PDFs) directly in tool result message - Auto-update and OSC9 bug fixes - Improve skills to mitigate potential model laziness

GPT-5.5 is so fucking good. I need sponsored credits.

@skcd42 And when faced with conflicting SPEC/docs, code, and user input the model(s) seem to be leaning toward what the code was recently rather that the vector of change, SPEC, or my specific instructions.

@skcd42 I also noticed some oddities in behavior that seem to be prompt induced — lots of noise about “this is just a simulation”, “I have no tools”.

Bug fixes shipping for Grok Build - Fix Windows contrast/color/theme rendering - Fix German QWERTZ AltGr on Windows - Convert session timestamps to local timezone - Add backslash continuation to plan mode - Fix auth for plugin-provided MCP servers - Default to PowerShell on Windows - Improve search tool on BM25 queries - Decrease large bash tool output to 20k chars front-and-back - Auto-install shell completions for bash, zsh, and fish - Force-refetch managed MCP configs - Don’t auto-wake the model on cancelled/killed tasks/subagents - Return images in tool response vs. deferred message

Ok let's dig in. Develop opinions, create confidence, execute.

@higgsfield not the first by a mile

Introducing Higgsfield Supercomputer The first ever cloud-native, self-learning AI agent for end-to-end task execution. 40+ built-in tools. Three layers of memory. Access via browser or Telegram. Powered by enhanced Hermes Agent.

here's all the Claude Code words in the style of Detroit Speed Rap

Qwen 3.5 397b a17b is cracked

Every day the DigitalOcean experience gets better and better.