Abhi Prajapati

@jssr @hosseeb @AnthropicAI ohh, could you please share the setup deets. I'd like to try it.

@abhip05 @hosseeb @AnthropicAI disagree. I'm doing presentations in custom design using CC that uses midjourney for backgrounds, and Gemini 3.1 Pro for text and diagrams.

update from the VC trenches: fucking purple accent lines are now in every single pitch deck thanks @AnthropicAI, i didn't realize how much i wanted everything to look like this

@simonw it was better when I tried it a few days back x.com/i/status/20375…

I'm a big fan of the pelican GLM-5.1 drew me today, it even animated it! simonwillison.net/2026/Apr/7/glm…

@ZixuanLi_ yesterday I was using it in cursor and it kept telling me that it couldn't access the tool call results even though cursor was showing it

If you’ve encountered garbled output like this while using GLM-5 or GLM-5.1 on our official service, the issue is now resolved. We've patched the underlying inference-side bugs and will be releasing an article shortly to dive into the technical details and our specific fixes.

hey @bcherny , i'm getting this error on claude code web even after it consumes usage limits for few minutes

is claude code on the web and app down ?

been using it in maxxclaw(my openclaw) since few days and it's really great at agentic tasks

@pedroenaus yeah, they've confirmed it : status.claude.com

@abhip05 Authentication error in my case

@hey_yogini yeah even gpt 5.1 codex mini is really good. I've been using it in my openclaw for a long time. and if anyone is looking for affordable ones with the same intelligence then all Chinese models are really good like glm-5, mimo v2, qwen 3.6 plus, minimax, etc.

Moved my OpenClaw from Claude to a mix of Gemini 3.1 Pro and Gemini 3.1 Flash-Lite. So far, the results are good. Seeing everyone else complaining about GPT 5.3, I think this was a good decision :)

@NASA wow !! the northern lights look beautiful from the space

1972 ➡️2026 Apollo 17 ➡️ Artemis II

@hey_yogini @AutoSendEmail this could also reduce the skill. md file context size when it needs to perform a sequence of tasks. cause now every API response itself could include the next and previous steps as suggestion on success or failure.

Our errors are now agent-readable. Coding agents needed a lot of context to figure out what to do from a plain error and status code. So today I am shipping structured error messages, built for agents. Every error now has five fields: message, code, status, details, and suggestions. Details tell exactly which field or value caused the issue. Suggestions point to the right docs to fix it. Saving tokens for you!

@hey_yogini @AutoSendEmail interesting!! I think every skill(especially for openclaw) should have this.

nice 😎 x.com/i/status/20299…

@Peerlist congratulations 🎉

200,000+ builders on Peerlist 🥳

@Ra1kshit @claudeai even a single prompt with few lines of response consumes around 5% since they've tightened the limits last week

Claude code is essentially unusable for me i hit 36% session limits in 15 mins. Just by using a single agent on my codebase. @claudeai you need to fix it. I'm on the 100$ plan btw

Session #2 is done! This time, we took @arbitrum Builder Pods to @SVNIT_Surat, Gujarat, one of the greatest engineering colleges in the region. The energy in the room was great, and the students came really well prepared. A quick recap below. 🧵

successfully passing the svgbench x.com/i/status/20299…

@reactive_dude claudemaxxing time 😅

touch grass time

@levelsio am I in the elite group 👀

Okay let's see who can reply to this

@Zai_org does it have same usage limits as GLM-5 ?

GLM-5-Turbo is now available for all GLM Coding Plan users.