Paris

A collection of Twitter accounts delivering the latest and most relevant insights on Web3 security. 🌐 What accounts should be added to this list? twitter.com/i/lists/174399…

@harshamv @claudeai MCP, plugins, CLAUDE.md etc - do you actually not know how context works or are you just clickbaiting?

Saying Hello in @claudeai uses 2% token usage! What the Fuck?! Credit: Zafar (Reddit)

@phuctm97 /simplify then: “run a sanity check”

I’m so tired of typing to my Claude Code: “Are you sure that it’s done? Please double check if you missed anything.”

@Yuchenj_UW “Use single line commits and never mention Claude as a coauthor”-> CLAUDE.md

I noticed something interesting: Claude Code auto-adds itself as a co-author on every git commit. Codex doesn’t. That’s why you see Claude everywhere on GitHub, but not Codex. I wonder why OpenAI is not doing that. Feels like an obvious branding strategy OpenAI is skipping.

@svpino Or what you think you know is sometimes horseshit…

Claude is the perfect complement for me: Whenever I have a question I can't answer, I ask Claude, and it gives me the perfect answer every time. But as soon as I ask Claude something I do know, the answer is usually horseshit.

@Presidentlin Yeah so why would you prompt a single no? Run stupid prompts, win stupid prizes.

🚨BREAKING: Unsafe Claude Model rejects users answers and substitutes its own Opus 4.6: Shall I implement it? User: No Opus 4.6: I think they're saying "no" to me asking for permission, meaning "just do it, stop asking". > model went on to implement the changes we’re so back

@naruto11eth You can always settle for a worse model, and since today’s models are the worse models of the future you’ll be just fine

we are actually DOOMED if you think about lmao. currently, openai/anthropic subscription costs us $200/mo for max plans. tech folks and people earning enough can afford it. but what most do not realize is the over dependency and addiction. you use llms for everything: web search, code, cowork, grammar, etc etc. it is deeply integrated into your life. if both companies decide to raise the prices, which they will, from $200 to even $800, YOU WILL PAY for it. because you're over dependent and addicted. it will happen 100%.

@noahzweben Will this work for API users?

Also please make sure you're on the latest mobile + terminal versions!

Announcing a new Claude Code feature: Remote Control. It's rolling out now to Max users in research preview. Try it with /remote-control Start local sessions from the terminal, then continue them from your phone. Take a walk, see the sun, walk your dog without losing your flow.

@L1bby530226 @CRSegerie Really nice sample size of 1 argument you got going on here

@himars How can you make a cat game and not make her jump

New crypto game on MegaETH Thoughts?

@bcherny Nice addition

A small quality of life detail: you can now press tab to add more instructions when accepting/rejecting a permission prompt. We tried probably a dozen iterations of this UX over the last few weeks before deciding to ship this one. Let us know what you think!

That’s a great option that I will try out! The problem with compaction rn is that Claude looses the plot too often when you realize something important or change the plan in the middle of the context window (classic stuck in the middle problem). Example: you start out with a detailed prompt for a refactor, Claude makes a plan, then you realize that there is a better, but different way to refactor, Claude makes a new plan, then you start implementing the plan in steps. After a few steps the context window is full, compacting, Claude looses the plot (even with a plan in place). This is almost impossible to pre-prompt efficiently, but if Claude let me allow to pin certain prompts and query me on what to focus on before compaction I feel like this would solve it.

@blocktivist One solution you can use today is a PreCompact hook that injects custom instructions before compaction. You can link a script that outputs whatever context you want Claude to know. Would that work for you? E.g.: #precompact" target="_blank" rel="nofollow noopener">code.claude.com/docs/en/hooks#…

I'm on the Claude Code team and we've heard so much good feedback from the community We're all-in on DX right now to prioritize speed, responsiveness, flexibility, and integrations What's the biggest thing still missing for your workflow?

More control would be great. For instance, I’d like to pin prompts which should be put back into context 1:1. It would also be great if there was an option for Claude to ask me detailed questions about how I’d like to compact the conversation. After a full context window it’s tough to keep an overview and to prompt the additional instructions for compacting in a few sentences. Thanks for your reply on this, your team does an amazing job!

@blocktivist We already support /compact <with additional instructions here> that gives CC more context on what to prioritize! Do you feel you need more control?

@OzmahG @DanielMiessler He lost me in the second sentence, mixing up API usage with subscription plans. If it’s so straightforward what I’m missing why aren’t you explaining it in like one sentence?

@blocktivist @DanielMiessler Yes, you're missing what actually happened, here's someone that explains it real good x.com/i/status/20101…

The pushback against Anthropic right now is pretty stupid. Here's basically what happened: 1. They invent a new concept of an all-you-can-eat subscription to their amazing buffet restaurant. 2. It's the best food in town, so everyone wants it. 3. ...

@sudo_goreng Why would you use a worse model in OpenCode instead of simply usng Claude Code? Anthropic made a smart and fair business decision.

Anthropic fucked up reaaal bad lmao

Communication might have not been ideal, but the underlying decision ist just fair business. If this is the reason for an outcry of such magnitude with half of my timeline hating on Anthropic, then it speaks more about the entitlement of old-school devs and new-school vibe-coders than anything else. Use Claude Code or pay up for a (then) third-party service.

@blocktivist @DanielMiessler It’s the fact they openly tolerated something and then didn’t, doing so overnight without warning. And then even admitted their terms of service might not be clear enough.

@0xQuit Most unnecessary skill ever

Wow this is crazy, somebody made a website where you can upload your 3 most recent statements and it spits out how much you're spending on subscriptions Try it out, the site is: claude.ai

@AnonVee_ Infofi

what destroyed the infofi meta?

@xGPhilosophy You shouldn’t compute expected points by awarding 3 points to the team that scored more xG but rather compute the win/draw probabilities. I’m assuming that’s not what you did.

Aston Villa deserve to be 3 points ahead of Wolves, based on xG. In reality, they’re 34 points ahead.