I don't understand why chatGPT and Claude don't allow you to queue messages or pass more in as they're responding like Claude Code, often find I want to refine the problem as I've remembered more context etc

This evening, I released a beta port of Superpowers and a Claude-style skills system for OpenAI's Codex CLI. blog.fsck.com/2025/10/27/s...

there should be a way to work for 20 companies at once and rotate them like once a month so you don’t get bored

Not sure if it was poorly advertised or I've just had my head in the sand but I didn't know Github now had an MCP registry github.com/mcp Not sure how they hand picked the 44 servers they've included, has a lot of the ones I'd use day to day although Linear is clearly missing

MCP Registry

A faster, safer way to build with AI: the GitHub MCP Registry centralizes MCP servers for effortless discovery, integration, and open collaboration.

github.com

Anyone see an improvement in copilot code completions since they “enhanced” the chatGPT 4.1 model the use? I know everyone is in cursor so hasn’t even tried it. github.blog/changelog/20...

GPT-4.1 Copilot code completion model — October update - GitHub Changelog

We’ve enhanced the GPT-4.1 Copilot code completion model in GitHub Copilot to improve the model’s ability to infer your intent from code context. As a result, you’ll see more accurate…

github.blog

@s.ly Do you know if you can run plugins inside Claude Code web? From the docs and experimenting with superpowers in it I'd say No. But its not immediate obvious and wondered if anyone else had tried

Started playing with obra/superpowers yesterday and my first impression is it's a step forward when working with Claude Code. Requires you to interact with the agent more but having a clear planning phase with specific outputs at each step really helps.

Good to see, cloudflare needs ti support popular frameworks to gain adoption of workers which is an underutilised choice IMO

Kenton Varda@kentonvarda.com · last yr.

Those benchmarks from @t3.gg revealed some interesting issues in Workers. Happy to say it's all fixed now, save for some lingering next.js-specific stuff we're continuing to work on. Thanks for the reproducible test cases, @t3.gg. Excessive details in blog post: blog.cloudflare.com/unpacking-cl...

Be careful out there devs, people will sell you a dream that doesn’t exist so you enrol in their $xxx course

🎯

The real rate limit is your credit card.

>Actually, I think I see the issue now. the biggest lie of all

I will die on this hill

Tom MacWright@macwright.com · last yr.

a directory called 'docs/' filled with markdown files is 1000% better for internal documentation than notion

junior dev: can you please give me an example of useful comments in my code? Me: sure, good comments look like this ⤵️

True story

This actually cost almost $15, if the output was decent it would have been money well spent but instead I paid $15 for trash.

Tim Reynolds@timjreynolds.bsky.social · last yr.

Will find out cost tomorrow once AWS billing catches up, when complete it said the work cost $5 but idk if it’s smart enough to know that via AWS costing.

Don’t want this to come across as another engineering being negative about AI and saying it won’t take my job. Engineers using AI are gaining super powers, those with experience are able to ship quicker or build more ambitious applications.

Tim Reynolds@timjreynolds.bsky.social · last yr.

Finally tried Claude Code today. Nice it just needs access to the API and support AWS and GCP endpoints out the box. Was easy to setup with the work AWS account to use Bedrock. Took a long time to gain context and perform the task of finishing our migration from one react state lib to another.

Finally tried Claude Code today. Nice it just needs access to the API and support AWS and GCP endpoints out the box. Was easy to setup with the work AWS account to use Bedrock. Took a long time to gain context and perform the task of finishing our migration from one react state lib to another.

The number one "leadership" trick has to be breaking something and then taking credit for fixing it.

The fact drive thru Greg’s exists tells you everything you need to know about the UK’s diet

Just published a highly opinionated guide to stand-up meetings. It’s based on ~25 years working with close to 100 product teams, and seeing first-hand what actually helps teams deliver - and what just wastes everyone’s time. If your stand-ups feel like a chore or status theatre, this is for you.

We're hiring a QA engineer at tldraw's London office. Help us wrangle our issues, hit our releases, and deliver a perfect product. Open source experience highly valued, deeply respected, quietly appreciated. jobs.ashbyhq.com/tldraw/a255...

Senior / Lead Quality Assurance Engineer

Quality Assurance Engineer @ tldraw

jobs.ashbyhq.com

I kind of want to go

After a decade I upgrade the MX Master 2S to the MX Master 3S and initial impression is good. Has a different click feel but the button placement tweaks are great

Not sure if its AI fatigue or if LLMs are going backwards but in the last 24 hours I've disabled Apple Intelligence and stopped using Cursor.

The mistake with Cursor is getting fixated on the agent when the real super power is the tab completion functionality

Personal post: My mum’s currently undergoing chemo, and five weeks today I’ll be running the London Marathon trying to raise £7500 for Macmillan I’m a LONG way from my target, so if you’ve ever found any posts helpful, please consider donating: 2025tcslondonmarathon.enthuse.com/pf/chris-med...