Skip to main content
Back

Anthropic presents Claude 4.5, best-in-class for coding + computer u

Shopping in ChatGPT, Lovable's AI full-stack apps, Microsoft brings agents to Office, Cursor adds browser control, Anthr

September 30, 2025 4 min read

Hey

Welcome to AlphaSignal, the most read source of news by AI engineers and researchers.


Every day, we identify and summarize the top 1% of news, papers, models, and repos, so you’re always up to date.


Here’s today’s roundup:

Summary

Read time: 4 min 17 sec

Top News

▸ Anthropic releases Claude Sonnet 4.5, sets new benchmark in coding tasks

AssemblyAI

▸ Build scalable Voice AI apps with AssemblyAI’s 93.3% accurate transcription

Signals

▸ OpenAI enables direct shopping inside ChatGPT

▸ Lovable combines Cloud and AI, enabling full-stack apps by prompt

▸ Microsoft brings agentic capabilities to Excel and Word

▸ Cursor introduces early preview of browser control with Sonnet 4.5

▸ Anthropic rebrands Claude Code SDK as Claude Agent SD

Trending Lectures

▸ Cursor launches Cursor Learn: starts with course on AI foundations

▸ OpenAI Academy releases ready-to-use prompts for daily workflows

▸ DeepLearning.AI's course on agentic coding course using Claude

AI Coding Tip

▸ Use adapters and feature flags to safely deploy Claude model upgrades

Top News

Anthropic’s Claude Sonnet 4.5 delivers benchmark leadership with enhanced coding and long-context performance

alpha_signal_image_2

Anthropic introduced Claude Sonnet 4.5, its strongest coding model to date. It outperforms previous versions in coding, reasoning, and math, and is tuned for building complex agents and extended computer use.


Key Features

  • Best-in-class coding: State-of-the-art on SWE-bench Verified.

  • Agent building: Optimized for orchestrating complex, multi-step workflows.

  • Extended use: Handles 30+ hours of sustained coding tasks.

  • Reasoning gains: Substantial improvements in math and logic benchmarks.

Performance

Claude Sonnet 4.5 shows the largest step-up across coding, reasoning, and applied benchmarks.

  • SWE-bench Verified: 77.2% (higher than earlier Claude models).

  • OSWorld (computer use): 61.4%, up from 42.2% in Sonnet 4.

  • Math (AIME 2025): 87.0% vs. 70.5% for Sonnet 4

  • Graduate reasoning: 83.4% vs. 76.1% for Sonnet 4

  • Benchmark leader: Now tops Anthropic’s coding and reasoning tests.

New Developer Tools

With this release, Anthropic also introduced several developer-focused updates:

  • Claude Code updates: Fresh terminal interface and new VS Code extension.

  • Checkpoints: Roll back instantly to previous states during large coding runs.

  • File handling: Claude can now create, edit, and visualize files in formats like spreadsheets, slides, and PDFs.

  • Chrome extension: Available to all waitlisted users for browser-native workflows.

API Upgrades

  • Context editing: Automatically clears stale context to extend session lifetimes.

  • Memory tool: Stores and retrieves information outside the context window for long-running tasks.

Access and Availability

  • Available today across the Claude Developer Platform, Amazon Bedrock, and Google Vertex AI.

  • Pricing unchanged from Sonnet 4: $3 / $15 per million tokens.

  • Also released in preview: Imagine with Claude, an experimental mode for on-the-fly software generation (5-day test, Max users only).

Try now
alpha_signal_image_3

93.3% Accurate Speech-to-Text to Build Voice AI at Scale

AssemblyAI gives you high-quality transcripts for Voice AI apps that scale. You pay only $0.15 per hour with no commitments.

  • No limits on concurrency, scale up instantly without throttling.

  • Reliable for downstream LLM and Voice AI workflows.

  • High-quality transcripts trusted by Granola, Cluely, and Dovetail

AssemblyAI improves models continuously, giving you faster shipping and trusted insights.

Start building for free

partner with us

Signals

OpenAI launches ACP to standardize programmatic commerce flows between AI agents and merchants

9,726 Likes

Lovable unifies backend and AI services, simplifying full-stack development into a single prompt

7,437 Likes

Microsoft debuts Agent Mode, enabling Copilot to iterate and validate tasks inside Excel and Word

2,354 Likes

Cursor adds browser debugging capability, letting agents inspect, improve, and test user interfaces

4,258 Likes

Anthropic introduces Claude Agent SDK, enabling agent loops, subagents, and flexible tool design

1,846 Likes

Lectures

Cursor launches Cursor Learn: stars with course on AI foundations

6,214 Likes

This is a six-part video course on AI foundations. It covers tokens, context, agents, and practical patterns for coding with AI tools. You learn limitations, workflows, and productivity strategies in under an hour.

OpenAI Academy releases ready-to-use prompts for daily workflows

5,673 Likes

OpenAI introduced role-specific Prompt Packs in its Academy, offering 300+ curated prompts for diverse jobs. These packs guide professionals in using AI tools effectively across technical, business, and customer-facing functions.

DeepLearning.AI's course on agentic coding course using Claude

3,957 Likes

The course teaches you how to use Claude Code to plan, write, test, refactor, and debug code with minimal human input. You’ll work on projects, RAG chatbot, notebook to dashboard, web app and learn best practices for context, multitasking, and integrations.

AI Coding Tip

Use adapters and feature flags to safely deploy Claude model upgrades

When integrating Claude Code, design implementations around interfaces. This makes it easier to swap in stronger models like claude-sonnet-4-5 without large rewrites.


What to do:

  • Wrap model calls in adapters

  • Control rollouts with feature flags

  • Keep key components interface-driven

Benefits:

  • Enables seamless upgrades as models improve

  • Reduces risk when deploying new versions

  • Provides flexibility across environments

Applications:
Gradual migration to Claude 4.5 Sonnet, production code rewrites, extending AI to daily workflows like forms, calendars, or reports.

Read now

At Alpha Signal, our mission is to build a sharp, engaged community focused on AI, machine learning, and cutting-edge language models, helping over 200,000 developers stay informed and ahead. We’re passionate about curating the best in AI, from top research and trending technical blogs to expert insights and tailored job opportunities. We keep you connected to the breakthroughs and discussions that matter, so you can stay in the loop without endless searching. We also work closely with partners who value the future of AI, including employers and advertisers who want to reach an audience as passionate about AI as we are.


Our partnerships are based on shared values of ethics, responsibility, and a commitment to building a better world through technology.Privacy is a priority at Alpha Signal. Our Privacy Policy clearly explains how we collect, store, and use your personal and non-personal information. By using our website, you accept these terms, which you can review on our website. This policy applies across all Alpha Signal pages, outlining your rights and how to contact us if you want to adjust the use of your information. We’re based in the United States. By using our site, you agree to be governed by U.S. laws.

Looking to promote your company, product, service, or event to 250,000+ AI developers?

Work with us