Mohamed Soliman

I built a website for people who hit their Claude usage limit. You describe what you were building. We match you with someone working on something similar. You chat until Claude comes back. And if we can't find you a match in 10 seconds? Claude roasts your idea instead. Simple. claudepause.com

I think I summoned something @ClaudeDevs

Anthropic with Software engineers:

Started a channel to yap about AI professionally instagram.com/solytalksai?ig…

software engineers before vs after agents

There’s something profound about sitting alone with your own thoughts and letting them unspool. Contemplation as its own reward. no output, no deliverable. just thinking, thoroughly

Coding harnesses are the clearest example of why the app layer / gpt wrappers/ harnesses matter. The same model using a different harness produces major differences in output quality. The intelligence is the same but the context, tools, prompt, skills, subagents. All of this architecture makes a difference. This wasn't obvious when everyone called any AI startup a wrapper 2 years ago. a harness is the same thing as a wrapper but sounds cooler. It was fair to call it a wrapper when you were just doing RAG. Now it's a material difference. Context engineering isn't trivial. Everything people were doing 6 months ago needed to be rebuilt: - CLI tools are more powerful than MCP - Skills let you add more context - Filesystem feels more natural for LLMs to work with Bad harness = stuff all the context in and hope the AI is smart enough to use what it needs. Expensive, slower, less accurate. Maybe the models will get good enough that you can afford to just throw everything in context. But that doesn't seem soon given how expensive inference is. At least the cost savings make a good harness worth it. The point is that the value is in the app layer. Everyone knows that now. The intelligence we have today can do a lot more with a 10x harness. We're still extremely early that a 10x harness improvement might be close. AGI might be a small step improvement in the models we have now but a big step improvement in the harness it's using.

watching Figma, a $10B UX design monopoly, melt down with a single Claude tweet.

Opus release ritual

me and @claudeai

Microsoft employees watching Anthropic ship better AI features for office than Copilot.

@RoundtableSpace You can go to claudepause.com as well

“Claude usage limit reached. Your limit will reset at 4 AM.” Me:

Enough internet for today I guess

Whoever made these emojis needs to be fired immediately.

Me prompting my coding agents at 3am

Transition to gpt in openclaw not going so great this far 😂

تهور واضح

My love language nowadays is sharing my CLAUDE.md files

Working on your own business is weird. Some days you feel like a genius. Other days you’re convinced you’re the dumbest person alive

Imagine being the president of Y Combinator and blocking every good engineer on this app because they pointed out you were using the wrong metric and instilling future generations with bad advice.

boris is such a good manager, the poor engineer that flubbed the map thing literally probably felt like dying after the leak happened, really delicate human situation to be in as a manager, that person could be tipping over the edge under the surface i know it's memes and code and money and blah blah but at a deep level i hope more people see the masterclass in management and looking out for your people that boris has done with this when you get to the point of managing people you hit this stark realization (if you're not a piece of shit) that you have a pretty big role in this person's day to day happiness, pretty cool to see him handle it so professionally and demonstrate high level leadership like that