Issue #14 | DailyPrompt

01 announcement

Anthropic Plants Its Flag in Sydney and Hires Snowflake Veteran to Run Australia and New Zealand

Anthropic officially opened its Sydney office on 27 April and named former Snowflake senior vice president Theo Hourmouzis as general manager for Australia and New Zealand. Sydney becomes the company's fourth Asia-Pacific outpost after Tokyo, Bengaluru and the soon-to-open Seoul base, and arrives alongside a slate of local enterprise wins with Commonwealth Bank, Quantium and several Australian research institutions.

Anthropic →

announcement

Xero Strikes Multi-Year Deal to Wire Claude Into Small Business Accounting

Xero unveiled a multi-year partnership with Anthropic that pipes Claude reasoning directly into the accounting platform via a new assistant called JAX, while letting customers pull their Xero data into Claude.ai for analysis and planning. The companies say financial information shared between the platforms is scoped to the session and is not used to train Claude.

Xero Blog

product

$ cat story_03.md

Claude Cowork Lands in Amazon Bedrock for Enterprise Roll-Outs

Anthropic and AWS made Claude Cowork available inside Amazon Bedrock, letting organisations deploy the desktop collaborator across knowledge workers while keeping inference inside their own AWS region. Pricing is consumption-based through existing AWS billing, and the deployment carries projects, artefacts, memory, connectors, skills, plugins and MCP servers, though Computer Use and the Skills Marketplace stay disabled because they require Anthropic-hosted inference.

$ open aws-machine-learning-blog →

product No. 04

Claude Code 2.1.121 Tackles Multi-Gigabyte Memory Leaks and Adds Always-On MCP Tools

The latest Claude Code release lands fixes for unbounded memory growth when handling many images, a separate two-gigabyte leak in the /usage screen and a Bash tool that broke when a working directory was deleted mid-session. New features include an alwaysLoad option that exempts MCP tools from search deferral, a claude plugin prune command, type-to-filter inside /skills and PostToolUse hooks that can rewrite tool output across every tool.

Source: GitHub →

announcement

AWS Pushes Anthropic Tie-Up Down to the Silicon With Trainium and Graviton

An AWS weekly roundup confirmed Anthropic is now training its most advanced foundation models on AWS Trainium and Graviton chips, with engineers from both sides co-designing directly with Annapurna Labs to squeeze more efficiency out of the silicon. The disclosure deepens the multi-billion-dollar capacity deal the two companies announced earlier in the month and tightens AWS's hold on Claude training and inference.

AWS News Blog →

community

Claude.ai Stumbles Through a Five-Hour Billing and Login Wobble on April 27

Claude.ai spent more than five hours in a degraded state on 27 April after an outburst of elevated billing errors stopped some users from upgrading or paying for subscriptions, with knock-on signup failures and tool-call problems. Anthropic's status page eventually flipped back to operational, but DownDetector showed user reports surging around mid-morning Eastern time before the company restored normal service.

Read on DesignTAXI Community →