Marco Marti
3.3K posts
Marco Marti
@marcodelic256
AI, building, Markets & Sound Founder | @FeelClearme
Tham gia Haziran 2012
747 Đang theo dõi255 Người theo dõi
@trq212 @ClaudeDevs Starting from being up and running would be awesome.
If then Opus stops getting constantly dumb and making constant mistakes that would be epic
English
we're doing a lot more of this, hunting down some of the most annoying bugs in Claude Code
let me know if you have any white whales
ClaudeDevs@ClaudeDevs
In the last four Claude Code CLI releases, we’ve shipped 50+ stability and performance fixes. Faster resume, stable auth, lower memory, fewer hangs: 🧵
English
@ClaudeDevs I wish that task would not require Codex to fix it 😫
English
Claude Code can now send push notifications to your phone when a long task finishes or Claude needs your input.
Walk away from the terminal, we'll let you know when it's done.
English
This clearly make you understand how 99% of AI users have no clue what they are talking about
Arena.ai@arena
GPT-5.5 by @OpenAI is now live in the Arena, landing across multiple leaderboards. Here’s how it ranks by modality: - Code Arena (agentic web dev): #9, a strong +50pt jump over GPT-5.4 - Document Arena (analysis & long-content reasoning): #6, on par with Sonnet 4.6 - Text Arena: #7, Math #3, Instruction Following: #8 - Expert Arena: #5 - Search Arena: #2 - Vision Arena: #5 Strong, well-rounded performance, especially in Code (+50 pts vs GPT-5.4). Congrats to @OpenAI on the release. Full category breakdowns by modality in the thread.
English
Anione actually using deepSeek?
DeepSeek@deepseek_ai
🚀 DeepSeek-V4 Preview is officially live & open-sourced! Welcome to the era of cost-effective 1M context length. 🔹 DeepSeek-V4-Pro: 1.6T total / 49B active params. Performance rivaling the world's top closed-source models. 🔹 DeepSeek-V4-Flash: 284B total / 13B active params. Your fast, efficient, and economical choice. Try it now at chat.deepseek.com via Expert Mode / Instant Mode. API is updated & available today! 📄 Tech Report: huggingface.co/deepseek-ai/De… 🤗 Open Weights: huggingface.co/collections/de… 1/n
English
Stop tweeting for a hot minute and update your Codex App to find full browser use, global dictation, non-dev mode, a new auto-review mode that is much safer than yolo, in-app docs and PDF viewer, and ... GPT-5.5.
OpenAI Developers@OpenAIDevs
With GPT-5.5, Codex now gets more of the job done across the browser, files, docs, and your computer. We've expanded browser use so Codex can interact with web apps, and test flows, click through pages, capture screenshots, and iterate on what it sees until it completes the task.
English
In the span of 10 minutes..
Claude admits that it regressed really badly and GPT 5.5 is released.
What an insane world
English
@bcherny @ReadySetBrian On xhigh it lasts like 10 min with the max sub
English
@ReadySetBrian Hmm are you seeing this with Opus 4.7 on xhigh effort and the latest version of Claude Code?
English
Canceled Claude max today, @bcherny whatever happened in the last 1-2 months is a significant regression.
The model feels like someone from OpenAI started working on trust and safety there. Opus thinking is significantly worse. Every statement is “here’s where I’d push back on that” and then proceeds to rattle off the most inane list of confused counter arguments.
It was perfect 3-4 months ago!!!
English
Opus 4.7 doing something silly
-Bigger problem than a missing slash command. You've got two distinct "orchestrator" concepts in parallel, and neither of us realized it when
we built the second one yesterday.-
Bro you're the one creating folders, not we
English
This makes the difference in how good your Business strategy will be
Marco Marti@marcodelic256
When you're brainstorming on something that is really important and complex with different files, reports, data and so on, a really good prompt that I found for Opus or GPT 5.4 XHigh is to say : now spawn a sub-agent, brief him on what you did and ask it to: review your work; propose a different angle; challenge you; propose fixes if needed.
English
When you're brainstorming on something that is really important and complex with different files, reports, data and so on, a really good prompt that I found for Opus or GPT 5.4 XHigh is to say :
now spawn a sub-agent, brief him on what you did and ask it to:
review your work;
propose a different angle;
challenge you;
propose fixes if needed.
English
@bcherny @HaneeefShiraz well with high I hit my limits in less than 1h with max sub..
I have recently changed from codex and I'm thinking to switch back
English
@HaneeefShiraz We increased rate limits for all subscribers, so you shouldn’t feel the increased token consumption
English
While it feels helpful, I just cannot shake off the feeling that this is how Anthropic can start to throttle legally.
1. Release new model and say 30% higher usage
2. Ask to use High to do what older models did much better just normal
3. Eventually folks realize they run out much faster
Sigh. What a pleasure it was coding using Opus 4.5. And where has it gone to now :(
Boris Cherny@bcherny
👋 Is there a specific issue you're hitting? If so, would you mind running /feedback and sharing the id here? That would be most helpful for debugging. There were a number of harness changes that may have caused this, all of which are fixed in the latest (last known issue was fixed in 2.1.116 today). We will be sharing more in a bit, and have also shared a few updated on X/Threads as we've been investigating. General tips: 1. Use Opus 4.7 + xhigh/max effort 2. Make sure you're using the latest version of Claude Code (currently 2.1.116)
English
basically the 100€ sub for Claude lasts as long as the 30€ OpenAI
English
Is Open AI manipulating the answers of its own model to boost their own adoption?
This happened several times when I ran tests and asked gpt to evaluate sonnet vs gemini flash vs gpt 5.4 mini.
Do you trust what a single model outputs?
English
Damn Anthropic killed it.
The new Opus 4.7 is insane. Going to move fro the 100$ to 200$ plan.
English
I hope that @AnthropicAI didn't use Opus 4.7 to build and ship their new desktop app.. it's so buggy.. unreal how bad Claude is for coding.
Don't get me wrong, it's still amazing for writing and for some business ideas but other than that coding is really like bad.
English
He's literally talking about taxing people who own property they don't live in in NYC, you don't have to pay the tax if you live there and the housing is so expensive that it could lower prices in Manhattan to where normal people could live there and not just Russian oligarch ghost condos
English