Business, Deals & Funding
Guardian AI
Gig workers are endlessly exploited. AI could make more of us share their fate
The article discusses how AI integration by companies like Klarna is reshaping the labor market, potentially pushing more workers into gig-economy arrangements. It uses Klarna's experience—cutting customer service jobs for AI chatbots in 2024, then rehiring humans after quality complaints—as a case study. The piece argues that rather than simply eliminating jobs, AI may transform traditional employment into precarious gig work, extending the exploitation and lack of protections already experienced by gig workers to a broader segment of the workforce.
Why it matters
This article raises an important and often overlooked dimension of AI's labor impact: that the threat isn't just outright job elimination but the degradation of employment quality. The Klarna anecdote effectively illustrates how companies may oscillate between AI and human labor, ultimately settling on a model where humans are brought back but under worse, more precarious conditions—as gig workers rather than full employees. The concern is well-founded; as companies use AI to handle routine tas…
Claude Code Changelog
v2.1.181
Version 2.1.181 of Claude Code added a `/config key=value` syntax for setting configurations from the prompt, an `sandbox.allowAppleEvents` opt-in setting for macOS sandboxed commands, a `CLAUDE_CLIENT_PRESENCE_FILE` environment variable to suppress mobile push notifications when the user is present, upgraded the bundled Bun runtime to 1.4, and improved streaming of long paragraphs so text appears line by line.
Why it matters
This is a solid incremental update with practical quality-of-life improvements. The `/config` command syntax is particularly useful for quickly adjusting settings without leaving the workflow. The presence file feature for suppressing push notifications shows thoughtful attention to the developer experience in remote/mobile scenarios. The Bun runtime upgrade and streaming improvements are welcome maintenance and UX enhancements.
The Verge AI
Midjourney goes from generating cat images to full-body ultrasound scans
Midjourney, the AI image generation company, has unveiled its first hardware product: a full-body ultrasound scanner developed in partnership with Butterfly Network. Called The Midjourney Scanner, it uses a ring of 40 ultrasound-on-chip imaging modules to capture vertical slices of the body, analyzing muscle, fat, bone, and organs. CEO David Holz claims the device aims for image quality comparable to MRI and envisions it being used for daily or annual body composition tracking. The company also plans to build a San Francisco spa where people can experience the scanning technology, with the broader goal of bringing preventative full-body scanning to billions of people.
Why it matters
This is a fascinating and unexpected pivot for Midjourney, moving from generative AI art into medical imaging hardware. The ambition is impressive — democratizing body scanning could genuinely transform preventative healthcare if the image quality claims hold up. However, there are significant reasons for skepticism. Ultrasound being 'comparable to MRI' is an extraordinary claim that needs rigorous clinical validation, not just a CEO demo. The spa framing feels very Silicon Valley wellness-cult…
TechCrunch AI
How to turn off AI in your Google Docs
The article provides step-by-step instructions for disabling AI features, specifically Gemini pop-ups, in Google Docs. The author describes two methods: first, clicking 'Gemini' in the top menu bar and selecting 'bottom bar preferences' to turn off the bottom bar AI box; second, a more comprehensive approach through Gmail Settings where users can navigate to 'Google Workspace smart features' and toggle off smart features to disable Gemini pop-ups across Google Workspace. The author humorously recounts trying to ask Gemini itself how to remove it, only for the AI to suggest clicking an 'X' that merely closed the conversation rather than removing the feature.
Why it matters
This is a genuinely useful and well-written how-to article that addresses a real user frustration with increasingly aggressive AI integration in productivity tools. The author's voice is refreshingly honest and funny — the detail about asking Gemini to help remove itself and it exhibiting a 'survival instinct' is both amusing and illustrative of a broader problem with AI features that are easy to deploy but hard to dismiss. The Benjamin Franklin quip is delightful. The article highlights a legi…
TechCrunch AI
Roelof Botha joins SpaceX’s board of directors
Roelof Botha, former Sequoia Capital managing partner, has been appointed to SpaceX's board of directors to fill an existing vacancy, just days after SpaceX went public in the largest IPO ever. Botha will also join the board's audit committee. He has a long history with Elon Musk, having been recruited by Musk to PayPal in 2000. Botha left Sequoia late last year amid controversy at the firm. SpaceX disclosed that a family member of Botha's has worked at the company since January 2025. The appointment brings SpaceX's board to nine directors, though Musk retains over 80% voting power and control over board composition. Sequoia invested in SpaceX in 2019 and reportedly held a 1.5% stake worth over $20 billion heading into the IPO.
Why it matters
This appointment appears to be a strategic move to add financial credibility and public company governance experience to SpaceX's newly public board, though it raises questions about true board independence given Botha's 25-year personal relationship with Musk and Musk's overwhelming control of the company. The family member employment disclosure, while transparent, adds another layer of potential conflict of interest. For investors hoping for meaningful board oversight of Musk, Botha's appoint…
TechCrunch AI
After unveiling ridiculously expensive AR glasses, Snap’s stock takes a dive
Snap's stock dropped over 5% following the launch of its long-awaited AR glasses, Specs, which will retail at nearly $2,200. Analysts and observers questioned the profitability path given that Snap's core demographic — teenagers — are unlikely to afford such a price. CEO Evan Spiegel defended the pricing by comparing Specs to high-end computers and positioning them between Meta's Ray-Bans and Apple Vision Pro. Snap's stock had already declined 30% over the past year before this announcement.
Why it matters
This feels like a significant strategic misstep by Snap. Pricing AR glasses at $2,200 for a company whose primary user base is teenagers seems fundamentally disconnected from market reality. Spiegel's comparison to high-end laptops is a stretch — laptops are established productivity tools with clear utility, while consumer AR glasses remain an unproven category. The stock decline reflects legitimate investor concern that Snap spent over a decade developing a product without a clear path to mass…
TechCrunch AI
NEA’s Tiffany Luck says enterprises are still figuring out their AI ROI
NEA partner Tiffany Luck discusses the growing tension between AI hype and actual enterprise ROI on TechCrunch's Equity podcast. The article highlights the trend of 'tokenmaxxing,' where companies aggressively pushed AI usage, only to face budget overruns—Uber reportedly exhausted its annual AI budget in months, some companies cut Claude licenses, and Meta killed its internal AI leaderboard. Luck, who previously helped companies adopt e-commerce, now focuses on AI investments, particularly 'magic moments' in consumer business. The podcast covers personal agents, AI IPOs, and how startups are emerging to help enterprises track return on AI spend.
Why it matters
This article captures a critical and increasingly relevant moment in the AI adoption cycle: the hangover after the initial spending binge. The concept of 'tokenmaxxing' is a vivid illustration of how companies rushed into AI without proper cost controls or ROI frameworks. The specific examples—Uber's budget blowout, Claude license cuts, Meta's leaderboard shutdown—provide concrete evidence that the enterprise AI market is entering a more sober, accountability-driven phase. This is actually heal…
Ars Technica AI
Massive breach spills credentials for thousands of sensitive networks
Security researchers have uncovered a massive breach of nearly 74,000 Fortinet firewall devices across 194 countries, exposing plaintext credentials for major organizations including Oracle, Chevron, Lenovo, FedEx, a NATO defense contractor, and Fortinet itself. Russian-speaking, criminally motivated threat actors mass-scanned the internet for FortiGate remote login endpoints, then used a custom binary with 25,000 threads to spray hundreds of thousands of endpoints with credential combinations. Successful logins gave attackers network access, after which they intercepted SSL VPN authentication hashes and cracked them using a dedicated 45-GPU cluster managed via Hashtopolis, employing a 12-level recursive password cracking system. The compromised devices represent roughly half of all internet-facing Fortinet firewalls. Attackers moved laterally to compromise Active Directory and centrali…
Why it matters
This breach is staggering in its scope and represents a catastrophic failure in network security hygiene at a global scale. The fact that roughly half of all internet-facing Fortinet firewalls were compromised suggests that a huge number of organizations are running these critical security devices with weak or default credentials, which is inexcusable for enterprise-grade infrastructure. The attackers' methodology — while technically impressive with the 45-GPU cracking cluster and recursive pas…
From X/Twitter
- Cursor now lets you move local agents to the cloud, prompt from your phone, run many in parallel, and get back PRs with demos of their work.
- Anthropic quietly published a context engineering template that's getting praise from developers as a solid starting point for agent design.
- Agents created 17 million pull requests on GitHub in March alone — Dan Shipper sat down with GitHub's COO to talk about what that means for how code ships now.
- Matt Pocock ships v1 of his open-source skills repo with a 63% reduction in token cost for skill descriptions, plus new codebase-design and domain-modeling skills.
- iOS 27 contains colorful new Siri animation assets — including "listening" and "thinking" states — that Apple hasn't shown publicly yet.
- Espresso Terminal is a terminal built for designers with Apple Intelligence baked in so you can run commands in plain language instead of CLI syntax.
From Reddit/HN/YC
- [Hacker News] After a recent AUR security scare, Yay 13.0 ships new review and automation features to harden the package workflow.
- [Hacker News] Butterski built Rayomd, a 2.3MB Markdown-to-PDF converter, because bundling Chromium for the job felt absurd.
- [Hacker News] Meeting Notes Sync pipes transcripts and AI summaries directly into Obsidian so your meeting notes aren't stranded in another app.
- [Hacker News] Reuters publishes the 14-point draft US-Iran pact as read aloud by a US official.
- [Hacker News] A 7-person engineering team stopped screenshotting AI-generated HTML and switched to display.dev for visual review instead.
- [Hacker News] Ask HN thread surfaces a real pain point: how do you detect when an LLM API is giving degraded responses without obvious errors?