So I spent most of yesterday going round in circles trying to sort out a technical problem. Around 5 hours in total. But working through it with Claude finally got me there, and I think it's a good example of where AI genuinely reduces cognitive load rather than just adding noise. The backstory: my clients push back on using Zoom. Some are blocked from it at work. So I finally made the move to Microsoft Teams. Three problems followed immediately. First, working out which Microsoft 365 licence actually matched my needs. Second, getting Teams to authenticate with two third-party platforms I use for scheduling. Third, and this is the one that nearly broke me: the error messages were completely useless. Just "something went wrong." No direction, no explanation, no next step. Claude walked me through the troubleshooting and eventually steered me toward entra.microsoft.com, a Microsoft admin portal I had never heard of and never logged into. Turns out some consent settings in there were blocking the integrations entirely. A few setting changes later, it worked. The point isn't that AI is magic. It took 5 hours and a lot of discussion and screenshots. The point is that having something to think out loud with, that could interpret vague error messages and suggest where to look next, made an otherwise impossible task manageable. And worth noting: professional IT support for this kind of issue is expensive and largely geared toward larger organisations on contracts. Small businesses and sole traders are mostly left to figure it out alone. AI doesn't replace expert technical knowledge, but it does level the playing field a little for those of us without an IT department.

This is what we ended up with after a non leading conversation. Even got Turn ons and Offs! lol

Microsoft 365 connectors are now available on every Claude plan. Connect Outlook, OneDrive, and SharePoint to bring your email, docs, and files into the conversation. https://x.com/claudeai/status/2040086268562842097?s=20

Happy Easter everyone! 🐣 Before you switch off for the weekend, came across this post on X that's worth bookmarking. The thing most people don't realise is that Claude counts tokens, not messages. So it's not about how many times you message, it's about how efficiently you use the context window. This post breaks down 10 habits that make a real difference: Use tokens wisely: - Edit your prompt instead of sending a correction - Batch multiple questions into one message instead of sending them separately - Start a fresh chat every 15-20 messages - summarise first, paste into a new conversation - Use Haiku for simple tasks like grammar, formatting, brainstorming - save the bigger models for real work Reduce what loads in the background: - Upload recurring files to Projects so they're cached, not re-tokenised every chat - Set up Memory and User Preferences so you're not burning 3-5 messages on setup every time - Turn off web search, connectors, and extended thinking when you don't need them Work with the system, not against it: - Spread work across the day - Claude uses a rolling 5-hour window, not a midnight reset - Avoid peak hours (1pm-7pm UK time on weekdays) when your limit gets consumed faster - Enable Extra Usage as a safety net - pay-as-you-go kicks in instead of blocking you Full post: https://x.com/0x_kaize/status/2038286026284667239 I'll be honest, some of these are harder to follow than they sound. They all rely on being switched on in the moment, and some days that just isn't available to me personally so I keep it simple - I don't run multiple Claude windows at the same time, and my working day is fairly short, so I've not hit the limits yet. Pick the habits that fit how you actually work and don't stress about the rest. Enjoy the break! 🐰

Today we're excited to announce NO_FLICKER mode for Claude Code in the terminal It uses an experimental new renderer that we're excited about. The renderer is early and has tradeoffs, but already we've found that most internal users prefer it over the old renderer. It also supports mouse events (yes, in a terminal). Try it: CLAUDE_CODE_NO_FLICKER=1 claude https://x.com/bcherny/status/2039421575422980329