Anthropic says the new Claude Sonnet 4.5 is the best coding model in the world, the strongest for building complex agents, and the best at using computers.

The launch post highlights gains in reasoning, math, and real computer control, plus a new Agent SDK and updates to Claude Code and the apps. Pricing matches the prior Sonnet 4. 

Anthropic cites a lead on SWE Bench Verified and a jump on OSWorld computer use, saying Sonnet 4.5 maintained focus for more than thirty hours on long tasks.

TechCrunch also reports enterprise trials where it coded for around thirty hours, stood up services, and handled compliance checks. These are early signals, but they point to longer horizon work that goes beyond simple code completion. 

Availability matters. Sonnet 4.5 is live in the Claude API and apps, and it is already in Amazon Bedrock, which gives teams a managed path with regions and enterprise guardrails.

That distribution can speed trials for companies that want agents but need established security, monitoring, and procurement paths. 

The claim is big. Real value will hinge on how it handles messy codebases, permissions, prompt injection risks, and long running plans. Anthropic says this is its most aligned frontier model yet, with work on prompt injection defenses and reduced sycophancy.

If those protections hold while the autonomy grows, teams may start shifting more glue work and multi step coding to agents, with humans steering design and reviews. 

OpenAI introduced Instant Checkout inside ChatGPT. People in the United States on Free, Plus, or Pro can buy directly from Etsy sellers today, with support for more than one million Shopify merchants coming soon. 

It supports single item purchases now, with multi item carts and more regions planned. OpenAI also published an open standard called the Agentic Commerce Protocol so developers and merchants can integrate with agent workloads.

The experience is simple. You find an item in chat, check its details, and confirm payment without leaving the thread. It is a small step on paper, but it removes a common drop off point between discovery and checkout. If Shopify support arrives on time, this could connect brands that already sell through Shop Pay with intent that begins in a chat prompt.

If chat becomes a place where browsing and buying happen together, it pressures the old funnel that relies on search result pages, affiliate lists, and retargeted ads.

The new protocol and the first integrations are early, but they signal where agent commerce may go next. Industry watchers frame it as a direct push into ecommerce that other assistants will need to answer.

⚡ DeepSeek V3.2 lands with price cuts across API

🚗 Tesla FSD v14 reportedly begins staged rollout

🗣️ Apple trials Veritas chatbot for Siri internally

🧠 Gemini 2.5 Flash gets faster preview builds

⚖️ Lexroom raises 19m for legal AI expansion in Europe

🧵 Huawei details open source AI platform roadmap

You’ve learned about ‘Instant Checkout’, now it’s your turn to use it for the first time!

Pick a 3-minute moment (e.g. morning coffee, commute, pre-bed scroll) and ask an AI (ChatGPT, Kling, etc.) to build that ritual for you.

Here’s what to do:

✍️ If it failed, consider what would have sped it up, or how you could have improved your prompting, and save it for next time.

Click below ⬇️

Will Claude Sonnet 4.5 keep the coding crown for long, or will a new contender take its place within days or weeks? And, what do you think about shopping inside of ChatGPT? We’d love to hear your thoughts! See you next time! 🚀

Stay informed, stay curious, and stay ahead with Jumble!

Zoe from Jumble