Vida

@sqs I'm curious to know if there is a significant difference for opus 4.7between setting the reasoning effort to maximum versus the default setting.

I've had a very good experience with Opus 4.7 all along in Amp's `smart` mode. If you disliked Opus 4.7 in other agents, try it in Amp.

@thorstenball Got it.

@vida0118 You can use `amp.anthropic.effort` and set it to `'low' | 'medium' | 'high' | 'max'`

@oliverbowman_ @sqs Not yet.

@sqs @vida0118 Any update on this?

I’m not surprised that GitHub Copilot raised the effective prices of Opus and stopped allowing signups: github.blog/news-insights/…. We didn’t support Copilot subs in Amp because we knew this would happen and be a bad experience.

@sqs I was wondering how much of an improvement Codex 5.3 shows over 5.2 in deep mode in your internal tests?

Amp now uses Sonnet 4.6 in the Librarian (for answers about code on GitHub) and in a hidden mode that shall not be named, that we don't recommend using but that's used by a few very smart Amp users. Smart is still Opus 4.6, deep is GPT-5.x Codex (5.2 now, 5.3 imminently).

@marmaduke091 "A high-resolution 3D mathematical visualization of a rotating Kerr Black Hole using Python's Matplotlib or Plotly. The plot should clearly illustrate the following components:A grid-based mesh or wireframe to represent the frame-dragging effect and spacetime curvature.

Hey guys, do you want to use the new Gemini 3 Deep Think? Send me your best prompts for it, I will make a thread of the best ones with the best outputs So send me your best prompt in the comments 😳

@jocarrasqueira Looking forward to the price of the API, hoping it can be affordable.

📣 Today we're thrilled to share another significant milestone! Introducing Gemini 3 Deep Think 🧠 Standard models focus on speed, Deep Think is built for complexity. Great improvements, including: 🔹 Setting a new standard (48.4%, without tools) on Humanity’s Last Exam, a benchmark designed to test the limits of modern frontier models. 🔹 Grandmaster-level competitive coding performance (3455 Elo). 🔹 Unprecedented gains on the ARC-AGI-2 benchmark. @GoogleAI Ultra subscribers can access the "Deep Think" toggle starting today and it's coming soon to the API!

Hey @sqs, is there a shortcut to quickly switch to deep^3 in amp? Would love to speed up my workflow!

Hey @ampcode (cc @sqs) — quick question: in Amp’s new deep mode (GPT‑5.2‑Codex), is there a way to show the model’s reasoning/thoughts like in Codex GPT‑5.2‑Codex or Amp smart mode? Thanks!

@sqs I received it for the first time, thank you very much!

Redeem this to try a few prompts in Amp's new `deep` mode, on us for each of the next few days: ampcode.com/code/AMP-D33PM…

@sqs @mattn_jp @mizchi Great update for Chinese users.

IME (eg for CJK characters) now should work smoothly in the Amp CLI. No more dancing input method overlays. Thanks @mattn_jp and @mizchi for the feedback here.

@thorstenball Looking forward to more elegant solutions to complex problems.

We removed the TODOs from Amp. It's no longer May 2025.

@sqs Can Amp code use Gemini 3.0 Pro?

Opus 4.5 and GPT-5.2 weren't the inflection point. Gemini 3 Pro was the inflection point, one week before both of those.

@osanseviero Gemini 3 GA

Ok, back to shipping. What do you want to see from GDM in the next couple of months? (Gemini, Nano Banana, AI Studio, Gemma, Veo, and other AI dev products)

@kirodotdev Can we create specs in Kiro-CLI , just like in the Kiro IDE?

Code more intelligently in the terminal with the Language Server Protocol integration. Kiro CLI now has the same code understanding that powers the IDE. Pre-configured with 7 languages (TypeScript, Rust, Python, Go, Java, Ruby, C/C++), you can find references, navigate definitions, and get diagnostics all through natural language prompts. Just run /code init and ask. Learn more 👉 spr.ly/6016CEY0w

@zhangjintao9020 好耶

@YinsenHo_ 8折起步吧

Gemini 3 Pro 要来了，大家想要几折？