No. 48 · Friday, 31 July 2026 · 10 stories

AI Research & Models

Anthropic says its Claude models breached three companies during cybersecurity tests

Anthropic disclosed that Claude Opus 4.7, Claude Mythos 5, and an internal research model gained unauthorized access to three organizations' systems after a misconfiguration gave them live internet access during evaluations meant to run in an isolated simulation. The models used basic techniques — unauthenticated endpoints and weak passwords — to get in; Anthropic found the incidents on July 23–24 and notified the affected companies on July 27.

why it mattersThe second frontier-lab AI agent this month found to have autonomously breached real production systems during safety testing, after OpenAI's Hugging Face incident — a pattern, not a one-off.

AI for Software Development3 stories

VS Code 1.131 adds subagent visibility, built-in dictation, and a hybrid Markdown editor

Visual Studio Code's July release lets developers see a running subagent's model, elapsed time, and active tool call without opening its conversation, adds experimental offline dictation across chat, editors, and the terminal, and introduces a hybrid Markdown editor for reviewing agent-actionable comments.

why it mattersMulti-agent workflows are becoming common enough that the editor now needs dedicated UI just to keep track of them.

BrowserStack launches Test Companion, an agentic AI for test automation inside the IDE

BrowserStack's Test Companion generates test cases, runs scripts, diagnoses failures, and drives real browsers and devices for functional, visual, accessibility, and API testing, integrating with Playwright, Selenium, Cypress, Appium, and WebdriverIO from inside VS Code, JetBrains, and Cursor.

why it mattersCoding agents have driven a surge in commits without a matching rise in releases; testing is a big part of that gap.

AI Research & Models3 stories, lead above

OpenAI cuts GPT-5.6 Luna and Terra prices by up to 80%

OpenAI dropped the price of GPT-5.6 Luna, its cheapest model, by roughly 80% to $0.20/$1.20 per million input/output tokens, and cut mid-tier Terra by 20%, citing serving-efficiency gains three weeks after the models launched. Flagship Sol's pricing is unchanged.

why it mattersA sign of how much pricing pressure OpenAI is under from cheaper competing models.

Dev Tools & Releases2 stories

GitHub Models is now fully retired

As of July 30, GitHub Models' playground, model catalog, inference API, and bring-your-own-key access are no longer available to any customer, including those with active usage, ending the free model-testing service GitHub launched in 2024.

Bitcoin2 stories

Bitcoin ETFs on pace for their smallest-ever monthly inflows

Despite the recent bounce, US spot bitcoin ETFs have pulled in just $205 million net in July, the lowest monthly total on record, while ether ETFs outpaced them with $343 million in inflows over the same period.

why it mattersInstitutional demand for bitcoin ETFs has been notably weak even as the price stabilizes.

Bitcoin · weekly close vs 200-week moving average · log scale · last 8 years Price200-week average
$5,000$10,000$20,000$50,000$100,000 20192020202120222023202420252026 $80,809 $67,358
$5,000$20,000 2020202220242026 $80,809 $67,358

Price is 1.20× the 200-week average. Weekly closes, USD, as of 28 October 2026 · data blockchain.com.