Business, Deals & Funding
Claude Code Changelog
v2.1.149
Version 2.1.149 introduces several improvements: the /usage command now shows per-category cost breakdowns (skills, subagents, plugins, MCP servers), the /diff detail view supports keyboard scrolling with various keys, markdown rendering now supports GFM task list checkboxes, a new enterprise managed setting 'allowAllClaudeAiMcps' was added for loading claude.ai cloud MCP connectors alongside managed-mcp.json, and a PowerShell fix was included (details appear truncated).
Why it matters
This is a solid incremental update with practical quality-of-life improvements. The per-category usage breakdown is particularly useful for understanding cost drivers, and the keyboard scrolling for diff views addresses a clear usability gap. The GFM checkbox rendering is a nice polish touch. The enterprise MCP connector setting shows continued investment in enterprise deployment flexibility. The truncated PowerShell fix note is slightly concerning as it leaves users guessing about what was act…
Guardian AI
Sunrise Movement takes credit for disrupting Trump’s New York state rally – as it happened
The article is a live blog covering multiple political developments: the Sunrise Movement climate activist group claimed credit for disrupting a Trump rally in New York state; US intelligence director Tulsi Gabbard is leaving her post after a difficult tenure; Secretary of State Rubio discussed the ongoing reduction of US troops in Europe, stating it has been a process since the start of the administration. The blog also references broader topics including AI, China, Iran, war powers, and the Strait of Hormuz.
Why it matters
This appears to be a typical Guardian live blog aggregating several significant US political stories from May 2026. The headline about the Sunrise Movement disrupting a Trump rally suggests ongoing activist resistance to Trump's political activities, while the mention of Gabbard's 'rocky tenure' as intelligence director and Rubio's comments about troop reductions in Europe point to consequential shifts in US national security and foreign policy. The combination of these stories paints a picture…
TechCrunch AI
AI is being used to resurrect the voices of dead pilots
The NTSB temporarily shut down public access to its docket system after people used AI tools to reconstruct cockpit voice recordings from a spectrogram image included in the accident docket for UPS Flight 2976, which crashed in Louisville, Kentucky. Federal law prohibits the NTSB from releasing cockpit audio recordings, but the docket contained a spectrogram—a visual representation of the audio data. Using AI tools like Codex along with the publicly available transcript, individuals created approximations of the dead pilots' voices and shared them online. The agency restored access to the docket system but kept 42 investigations closed pending review, including the Flight 2976 case.
Why it matters
This is a genuinely disturbing development that sits at the intersection of AI capability, privacy, and the ethics of death. The spectrogram-to-audio reconstruction represents a foreseeable but poorly anticipated vulnerability—the NTSB followed the letter of the law by not releasing audio, but the spectrogram was functionally equivalent. It's a stark reminder that in the AI era, the distinction between 'data' and 'the thing the data represents' is collapsing. The voices of pilots in their final…
TechCrunch AI
Google goes for the glitter with disco-ball icons: ‘Are y’all sure you still want this?’
Google released a set of disco-ball-themed app icons for Pixel phones, joining in on the viral reaction to Spotify's temporary disco ball icon that celebrated its 20th anniversary. Android ecosystem head Sameer Samat announced the playful icon pack on X, asking users if they really wanted it. The icons are available through Pixel's custom icons feature, which launched in March's Pixel Drop and uses AI-generated styles to customize app icon appearances. The move was largely seen as a lighthearted joke, with mixed reactions ranging from amusement to mock horror.
Why it matters
This is a fun, low-stakes marketing move by Google that shows a willingness to be playful and culturally responsive. It cleverly capitalizes on the viral backlash to Spotify's disco ball icon while also showcasing Pixel's custom icon feature, which many users might not have known about. It's the kind of self-aware, meme-friendly corporate behavior that actually works — it doesn't take itself too seriously and gives users an optional, silly customization. Smart PR wrapped in glitter.
TechCrunch AI
How VCs and founders use inflated ‘ARR’ to crown AI startups
The article exposes a widespread practice among AI startups of inflating their publicly reported Annual Recurring Revenue (ARR) figures, often by substituting Contracted ARR (CARR) — which includes revenue from signed but not yet onboarded customers — and simply calling it ARR. Spellbook CEO Scott Stevenson sparked broad discussion by calling this a 'huge scam,' and TechCrunch's investigation with over a dozen founders, investors, and finance professionals confirmed the practice is common. Sources noted that investors are often fully aware of the exaggerations, and one VC reported seeing companies where CARR was 70% higher than actual ARR, with significant portions of that contracted revenue unlikely to materialize. The practice creates competitive pressure, as startups feel compelled to inflate metrics when competitors do so. Bessemer Venture Partners has noted that CARR should be adju…
Why it matters
This article highlights a genuinely important and corrosive problem in the AI startup ecosystem. The inflation of ARR metrics is not just misleading PR — it distorts capital allocation, creates unfair competitive dynamics, and undermines trust in the entire venture ecosystem. When major VCs knowingly support inflated metrics for portfolio companies while journalists uncritically report them, it creates a feedback loop that rewards dishonesty over genuine business building. The fact that startup…
The Verge AI
Google’s AI search is so broken it can ‘disregard’ what you’re looking for
Google's AI Overviews feature in search is malfunctioning when users search for action-related terms like 'disregard,' 'ignore,' and 'skip.' Instead of providing relevant search summaries, the AI interprets these words as instructions and responds like a chatbot, saying things like 'Got it. If you need anything else, just let me know!' Google acknowledged the issue, stating they are aware AI Overviews are misinterpreting action-related queries and are working on a fix. As of the article's writing, Google stopped showing AI Overviews for 'disregard' but the problem persisted for other terms like 'ignore' and 'skip.'
Why it matters
This is a revealing and somewhat embarrassing bug that highlights a fundamental architectural problem with Google's AI Overviews: the system apparently cannot distinguish between a user's search query and a conversational instruction directed at the AI itself. It suggests the underlying model powering AI Overviews retains chatbot-like instruction-following behavior that wasn't properly constrained for the search context. While the individual instances are humorous, they raise serious questions…
TechCrunch AI
Elon Musk can’t hear you over the sound of his $1.75 trillion IPO
TechCrunch's Equity podcast discusses SpaceX's S-1 filing for what would be a $1.75 trillion IPO — potentially the largest in American history. The filing details 36 pages of risk factors, a $28 trillion total addressable market claim, and an executive pay package tied to establishing a Mars colony. The episode also covers NanoCo turning down a $20 million buyout to raise a $12 million seed round, Anthropic's $300 million acquisition of SDK startup Stainless, negative student reactions to commencement speakers promoting AI, and Google's I/O announcements about AI transforming search and its implications for the open web.
Why it matters
This article highlights the extraordinary scale of SpaceX's IPO ambitions and Elon Musk's continued ability to command staggering valuations. The $28 trillion TAM claim and Mars-colony-linked compensation package seem designed to stretch investor imagination far beyond current financial reality. While SpaceX has genuine achievements in rocketry and Starlink, tying executive pay to Mars colonization in a public offering filing raises serious questions about whether this is visionary ambition or…
Ars Technica AI
Texas AG sues Meta over claims that WhatsApp doesn't provide end-to-end encryption
Texas Attorney General Ken Paxton has sued Meta alleging that WhatsApp does not actually provide the end-to-end encryption it claims. The lawsuit's primary evidence is a Bloomberg article reporting that a US Commerce Department investigation into Meta's WhatsApp encryption practices was abruptly closed after a preliminary finding email suggested Meta could access encrypted messages. However, the AG's office does not appear to have obtained the email itself or gathered independent evidence. The complaint also references Meta employees receiving plaintext WhatsApp messages reported by users, though these are decrypted on the reporting user's device before being sent. Meta called the allegations baseless. Encryption experts and technologists have criticized the lawsuit for lacking factual support, noting that reverse engineering of WhatsApp would likely reveal any bypass of the Signal prot…
Why it matters
This lawsuit appears to be more political theater than a serious legal challenge. The fact that the sole factual basis is a Bloomberg article about a closed investigation, rather than any independent technical analysis or firsthand evidence, is a glaring weakness. WhatsApp's implementation of the Signal protocol has been scrutinized by independent researchers and cryptographers, and the distinction between E2EE being broken versus users voluntarily forwarding decrypted messages to Meta through…
From X/Twitter
- CasJam is building internal tools for a content pipeline with a customer base of one — himself and his agents.
- Aakash Gupta says most people wipe Claude's memory daily and wonder why it never improves — the fix is persistent context that carries over between sessions.
- 500 founders, $10M on the table — the same crowd that mocked AI agents is now pitching agent-first startups by Friday.
- Polsia raised $30M at a $250M valuation with zero employees — one founder plus AI, approaching $10M ARR, and the AI even ran its own fundraise.
- Perplexity Computer now uses aggregated Ramp credit card spend to show real-time SaaS adoption, churn, and market share data.
- A Microsoft senior AI developer showed how they build agents with Claude Opus 4.7 and 1,400+ pre-built MCP tools inside Microsoft.
From Reddit/HN/YC
- [Hacker News] Pip 26.1 ships dependency cooldowns and experimental lockfile support.
- [Hacker News] A tool that source-checks politician stock-trade claims against actual public filings.
- [Hacker News] An interactive linear algebra primer designed not for humans but for LLM readers.
- [Hacker News] A couple found a baby on the subway — now he's their 26-year-old son.
- [Hacker News] Only 17% of all 64-bit integers are products of two 32-bit integers — Daniel Lemire does the math.
- [Hacker News] Sean Goedecke makes the case that the "just say no" engineer was a ZIRP phenomenon — and that era is over.