Business, Deals & Funding
Claude Code Changelog
v2.1.160
Version 2.1.160 of Claude Code adds security prompts before writing to shell startup files and git config directories to prevent unintended command execution. The acceptEdits mode now also prompts before writing to build-tool config files that could grant code execution (like .npmrc, .yarnrc, bunfig.toml, .bazelrc, .pre-commit-config.yaml, .devcontainer/, etc.). Additionally, the Edit tool no longer requires a separate Read operation after viewing a file with grep - single-file grep/egrep/fgrep commands now satisfy the read requirement.
Why it matters
These are solid, security-focused improvements. The added prompts before writing to shell startup files and build-tool configs address real attack vectors where an AI agent could be tricked into modifying files that execute code implicitly. The grep-satisfies-read optimization is a nice quality-of-life improvement that reduces unnecessary file operations. Overall, this is a thoughtful incremental release focused on tightening the security model of an AI coding agent.
Claude Code Changelog
v2.1.161
This patch release (v2.1.161) of Claude Code includes several improvements: OTEL_RESOURCE_ATTRIBUTES values are now included as labels on metric datapoints for slicing usage metrics by custom dimensions; the `claude agents` command now shows progress (done/total) with peek showing the longest-running item; the `/mcp` command collapses unused claude.ai connectors behind a toggle; and parallel tool calls are improved so that a failed Bash command no longer cancels other calls in the same batch.
Why it matters
This is a solid incremental release with practical quality-of-life improvements. The OTEL metrics labeling enhancement is particularly useful for teams tracking usage across different dimensions. The parallel tool call fix for failed Bash commands is an important reliability improvement — previously having one failure cancel sibling calls was a significant limitation. The MCP connector UI cleanup and agents progress display are nice UX touches that reduce noise and improve visibility into runni…
Guardian AI
UK media groups given power to opt out of Google AI search summaries
The UK competition watchdog has announced new conduct requirements giving media publishers the ability to opt out of having their content used to train Google's AI models and power AI search summaries such as AI Overviews. The regulator stated that publishers will now have effective tools to prevent their content from being used in AI-powered search features.
Why it matters
This is a significant and welcome development for press freedom and publisher rights. For too long, large tech platforms have been able to scrape and repurpose journalistic content to generate AI summaries that effectively replace the need for users to visit the original source, undermining the economic model that sustains quality journalism. Giving publishers an opt-out mechanism is a reasonable middle ground — it doesn't ban AI search features outright but restores some agency to content crea…
Guardian AI
‘The CGI would have cost millions. I spent $2,000.’ Is Dreams of Violets AI slop – or the future of film-making?
Dreams of Violets, a 75-minute drama about Iran's anti-government protest crackdowns, is set to premiere at the Tribeca Film Festival, making it the first AI-made movie to screen at a major film festival. Created by director Ash Koosha for approximately $2,000 in weeks rather than years, the film is based on journalism, video footage, and eyewitness accounts. The production raises questions about whether AI-generated films represent the future of independent filmmaking or constitute low-quality 'AI slop,' with Koosha arguing the technology could transform indie cinema by dramatically reducing costs and production timelines.
Why it matters
This represents a genuinely significant moment in the intersection of AI and filmmaking. The fact that a major festival like Tribeca is screening an AI-generated feature signals institutional legitimacy that could accelerate adoption. However, the tension captured in the headline — 'AI slop or the future of filmmaking' — is the right question. The $2,000 budget versus millions in CGI costs is compelling for indie filmmakers, but cost reduction alone doesn't validate artistic quality. The subjec…
Guardian AI
Sydney academic used AI to write SMH opinion piece urging students to avoid using tech to ‘cut corners’
A Sydney academic, Cath Ellis from Western Sydney University, used AI to write an opinion piece published in the Sydney Morning Herald that urged students to 'do the work' and avoid using AI technology to cut corners. The Sydney Morning Herald deemed the piece 'unacceptable' and removed it from its website after the AI-assisted authorship was revealed, though Western Sydney University defended her use of AI as 'appropriate.'
Why it matters
This is a striking case of hypocrisy that undermines the credibility of the message and the messenger. An academic urging students not to use AI to cut corners while doing exactly that to produce the very piece delivering that message is deeply ironic and damaging to public trust in academic integrity standards. While there may be nuanced arguments about 'appropriate' uses of AI in professional writing versus student assessment, the optics are terrible and the SMH was right to remove the piece.…
Guardian AI
Chip, chip ... boom? South Korea tech makers join the trillion-dollar club but some fear a short-circuit looms
South Korea's Kospi stock market has reached record highs driven by the AI boom, with the country leapfrogging India to become the world's sixth largest share market. Chip company SK Hynix and presumably Samsung have joined the trillion-dollar club. However, experts are raising concerns about the market's heavy dependence on just two chipmaking companies and warn about potential boom-bust cycles, suggesting the rally may be vulnerable to a correction.
Why it matters
This situation echoes classic market concentration risks we've seen before, such as the dot-com bubble or even the recent Magnificent Seven dominance in US markets. While South Korea's semiconductor companies are genuinely world-class and AI demand is real, having a national stock market so heavily dependent on two companies in the same sector is inherently fragile. Any downturn in AI chip demand, geopolitical disruption, or technological shift could disproportionately impact the entire Korean…
NY Times
Scientists Find Way to Supercharge Dangerous Computer ‘Worms’ With A.I.
Researchers at the University of Toronto have demonstrated how artificial intelligence can be used to supercharge computer worms, enabling hackers to create programs capable of automatically targeting any known vulnerability in computer systems worldwide. The research highlights a significant escalation in potential cyber threats through the combination of AI and existing exploit techniques.
Why it matters
This is a deeply concerning but important piece of research. While the publication of such findings always raises dual-use concerns — potentially giving malicious actors a roadmap — responsible disclosure of these capabilities is crucial for the cybersecurity community to prepare defenses. The fact that AI can be leveraged to automate and scale the exploitation of known vulnerabilities underscores the urgent need for organizations to patch systems promptly and for the security industry to devel…
TechCrunch AI
Cyera eyes $12B valuation at 80x ARR multiple despite operating losses
Cyera, a data storage security company founded in 2021, is finalizing a funding round of at least $300 million at a $12 billion valuation, led by Evolution Equity Partners. This values the company at 80 times its annual recurring revenue (ARR), which has surpassed $150 million. Despite rapid revenue growth — more than tripling in 2025 — Cyera remains far from profitable and is spending faster than it earns, partly due to aggressive hiring (500 new jobs this year) and acquisitions of startups like Ryft and Genie Security. The round comes just five months after a $400 million Series F at a $9 billion valuation led by Blackstone, bringing total capital raised to at least $2 billion. Cyera claims one-fifth of the Fortune 500 as customers, benefiting from enterprise demand for AI-era data security. The company's spokesperson disputed the reported financial figures as 'factually and significa…
Why it matters
An 80x ARR multiple for a company that is burning cash faster than it earns it is a striking signal of just how frothy cybersecurity valuations have become, particularly for companies positioned at the intersection of data security and AI threats. While Cyera's growth trajectory is genuinely impressive — tripling revenue and landing Fortune 500 clients — the fundamentals raise serious questions. Raising $2 billion in total capital while remaining unprofitable, acquiring young startups, and hiri…
From X/Twitter
- Claude Opus 4.8 + Higgsfield MCP finds ugly Airbnb listing photos and turns them into cinematic video tours in a single prompt — a service play for hosts and realtors.
- You had 400 good ideas this year and remember 6 — the case for Obsidian's local-first plain .md files over cloud apps you'll never reopen.
- Anthropic released its internal Prompting Playbook for free — 33 minutes on why most prompts fail, how top AI teams test outputs, and when to hand off to humans.
- OpenAI Codex ships 62 apps and 110 skills across six role-specific plugins — from data analysis to banking — plus a Sites feature that goes from idea to shared URL in minutes.
- Most people who download Claude never find Cowork. Here's the four-move system for turning it into a persistent workspace, not a one-off chatbot.
- Mario Zechner recommends reading up on dynamic workflows with durable execution — says the implementation is smart despite a few minor footguns.
From Reddit/HN/YC
- [Hacker News] New York is considering a mandate requiring 3D printers to include blocking technology — consumer rights wiki has the details.
- [Hacker News] Microsoft's Project Solara is a dedicated OS designed to run AI agent gadgets.
- [Hacker News] AgentCAD lets Claude Code drive a full CAD tool — built as a Show HN.
- [Hacker News] Brennan Day makes the case that the internet needs more cross-pollinators bridging disconnected communities.
- [Hacker News] MIT's 1986 Structure and Interpretation lectures are still worth watching in full.
- [Hacker News] UK publishers can now opt out of Google AI search results — BBC reports.