Business, Deals & Funding
Guardian AI
One of science fiction’s greatest writers warned us about a AI. Does he also hold the remedy? | Alan Finkel
The article discusses how Isaac Asimov's laws of robotics could be adapted for modern AI governance. Author Alan Finkel proposes 'three laws of AI' inspired by Asimov's framework, referencing Elon Musk's warnings about AI-powered robots potentially refusing human orders, while also noting Musk's alternative vision of making AI benign by instilling it with truth-seeking and pro-humanity values.
Why it matters
This is a thoughtful piece that draws on a well-established science fiction tradition to address real and pressing concerns about AI safety and governance. The idea of updating Asimov's laws for the AI era is a compelling framing, though the practical challenge has always been translating elegant principles into enforceable technical constraints. Musk's dual messaging—warning of dystopia while proposing utopian solutions—is characteristic but worth engaging with seriously. The article appears t…
Guardian AI
Can the government really get ahead of the curve on AI? - podcast
The Australian Labor government announced environmental and energy safeguards for new data centres. Assistant minister Dr Andrew Charlton discussed national data centre standards, his experience having his book scraped by Anthropic, low public trust in AI, and the government's approach to balancing regulation with risk.
Why it matters
This podcast touches on critical and timely issues around AI governance, data centre environmental impact, and the tension between innovation and regulation. The mention of Charlton's personal experience with AI scraping adds an interesting dimension. However, the question of whether any government can truly get ahead of the curve on AI remains deeply challenging, and it will be important to see whether these safeguards amount to meaningful action or symbolic gestures.
TechCrunch AI
OpenAI’s new AI smart speaker will reportedly sell for between $300 and $400
OpenAI is developing an AI-powered smart speaker in partnership with Jony Ive's design studio LoveFrom, expected to retail for $300 to $400. The device is described as donut-shaped, made from high-quality metal with a premium look and mysterious 'moving parts,' designed to be portable within the home. It is intended to be the 'physical manifestation' of ChatGPT and is expected to launch sometime in 2027. The price point is notably higher than most Amazon smart speakers, which range from $40 to $240. OpenAI is also currently being sued by Apple over allegations of stealing trade secrets, which the company denies.
Why it matters
This is a bold but risky move for OpenAI. Entering the hardware market is notoriously difficult, and smart speakers have historically struggled with profitability — even for established players like Amazon, which reportedly lost billions on Alexa-related devices. The $300-$400 price point puts it in premium territory without the established ecosystem that justifies such pricing for products like Apple's HomePod. The Jony Ive partnership adds design credibility, but a beautiful device alone won'…
The Verge AI
Jony Ive’s first OpenAI gadget is reportedly a hockey puck-sized smart speaker
OpenAI and former Apple designer Jony Ive are developing a battery-powered, doughnut-shaped smart speaker roughly the size of a hockey puck, expected to launch in 2027 for over $300. The device will have no display but will feature moving parts, lights, a camera system, and sensors. It is designed to be portable and will function similarly to ChatGPT voice mode but with more advanced conversational AI. It could be the first in a planned 'family of devices' from OpenAI.
Why it matters
This feels like another entry in the growing graveyard of AI hardware gadgets trying to find a use case beyond what a smartphone already does. At over $300 for a screenless speaker with ChatGPT voice mode, it's hard to see what compelling value this offers over existing smart speakers or just using the ChatGPT app on your phone. Jony Ive's design pedigree and the moving parts concept are intriguing, but the Humane Pin and Rabbit R1 have already shown that novel AI hardware without a clear kille…
Guardian AI
Safety fears as scientists make first viruses designed by AI
Scientists have created the first viruses designed by artificial intelligence, specifically bacteriophages that infect bacteria and are used to treat persistent infections. The breakthrough offers potential for new medicines, as AI-designed phage cocktails showed effectiveness in lab tests. However, the development raises significant biosecurity concerns about ensuring the technology remains safe and is not misused.
Why it matters
This represents a genuinely significant dual-use technology milestone. The therapeutic potential of AI-designed bacteriophages is exciting, particularly for combating antibiotic-resistant infections, which are a growing global health crisis. However, the biosecurity implications are serious and deserve careful attention. The fact that AI can now design functional viruses means the barrier to creating novel biological agents has been lowered, and governance frameworks need to catch up quickly. T…
The Verge AI
Suno shares plans to combat spammy AI music
Suno announced plans to implement new watermarking technology, fingerprinting tools, and a revised download policy to combat the spread of spammy AI-generated music and increase transparency. CEO Mikey Shulman outlined the company's principles in a blog post, emphasizing partnerships with distribution platforms to fight fraud and misuse. The changes follow settlements with Warner Music Group, which also led competitor Udio to end downloads entirely. Suno plans to limit downloads to paying subscribers with monthly caps. Shulman stressed transparency and respect for rights holders while maintaining the company won't judge the value of individual pieces of art.
Why it matters
This feels like a necessary but somewhat belated move by Suno to address legitimate concerns about AI-generated music flooding platforms with low-quality spam. The watermarking and download restrictions are practical steps, but the framing feels carefully crafted to appear responsible while still preserving Suno's core business model. Shulman's language about not 'passing judgment' on art's value is a convenient dodge — the real issue isn't artistic merit but the mass generation of content that…
TechCrunch AI
ChatGPT brings unlimited text chats to free users
OpenAI is removing limits on text-based chats for all free and Go tier ChatGPT users, powered by the new GPT-5.6 Luna model replacing GPT-5.5. Free and Go users will also get a 'Think' button for complex queries, though limits remain for files, images, voice, and image generation. Plus and Pro users receive an upgraded GPT-5.6 Sol model optimized for quick tasks with more compact answers, along with a thinking slider to adjust reasoning depth. OpenAI claims factual errors are 62% less common with GPT-5.6 Luna and 68% less common with GPT-5.6 Sol compared to GPT-5.5-Instant. The Sol model is available to Plus and Pro users immediately, with free tier changes rolling out over the following week. ChatGPT recently crossed 1 billion weekly users.
Why it matters
This is a strategically significant move by OpenAI as it continues to aggressively expand its free tier to maintain dominance in the consumer AI space, especially after hitting the 1 billion weekly user milestone. Removing text chat limits for free users effectively eliminates one of the biggest friction points that might push casual users toward competitors like Google's Gemini or Anthropic's Claude. The tiered thinking controls (a simple button for free users vs. a granular slider for paid us…
TechCrunch AI
Naïve raises $28.5M to automate the grunt work of setting up and running a company
Naïve has raised $28.5M in a Series A led by Nexus Venture Partners for its infrastructure platform that lets AI agents automate the process of setting up and running a business. The platform packages company incorporation, payments, email, phone numbers, cloud infrastructure, and connections to services like Stripe and QuickBooks behind a single API. Developers use tools like Cursor, Claude Code, or Codex with Naïve's prompt to provision business infrastructure, though humans are still required for KYC/KYB and payments. The company has signed up over 30,000 developer customers within months of launch and scaled annual run-rate revenue 10x to low double-digit millions over six months. Customers use it to run autonomous businesses including AI automation agencies, faceless TikTok/YouTube content channels, and even a rental car agency. The platform includes a governance layer for budgets,…
Why it matters
This is a fascinating and somewhat unsettling development. On one hand, Naïve is solving a real pain point by abstracting away the tedious operational setup of running a business, and the traction numbers are impressive. On the other hand, the use cases described — faceless AI-generated content channels, fully autonomous businesses — raise serious questions about the quality and authenticity of what's being produced, and whether this accelerates a race to the bottom in certain markets. The gove…
From X/Twitter
- Kris Puckett shows off the Paper MCP integration with Claude — prompt in, formatted research out.
- A sleep doctor with 26 years of experience explains why 8 hours is a myth, melatonin is misunderstood, and everyone wakes at 3AM.
- Keychron is prototyping a tiny wireless AI companion for vibe coding — built-in mic, scroll wheel, three programmable buttons.
- Mark Gurman reports OpenAI's hardware device will be a doughnut-shaped smart speaker the size of a hockey puck — camera, moving parts, $300–400.
- DSPy's new Flex idea lets optimizers replace AI calls with deterministic code — sometimes beating GEPA on both accuracy and cost.
- Pawel Huryn open-sourced 110 Claude skills for product managers — the top 30 have racked up 24K GitHub stars.
From Reddit/HN/YC
- [Hacker News] The Austrian military explains why and how it moved to LibreOffice — a detailed migration playbook in part one of the series.
- [Hacker News] Running DeepSeek V4 Flash 0731 locally vs. paying per token — a cost breakdown for self-hosters weighing the math.
- [Hacker News] A developer asks HN: after 20 years of coding it's not a fun hobby anymore — and the replies hit a nerve.
- [Hacker News] INT2 KV-cache quantization is getting surprisingly good — new results suggest aggressive compression may finally be practical.
- [Hacker News] A new analysis makes the case that the AI slowdown is coming — and the reasons aren't what the hype cycle wants to hear.
- [Hacker News] Every byte of memory manufactured in 2027 has already been sold out as the capacity crunch rolls into another year.