Good morning, AI enthusiast.

A prominent 'AI' startup, once valued at $1.5 billion and backed by major names including Microsoft, has reportedly collapsed. Allegations suggest its advertised AI capabilities were significantly overstated, with the work primarily done by human developers rather than actual artificial intelligence.

This development, apparently triggered after a key lender withdrew significant funds, brings to light long-standing questions about the company's technological claims. What might this unraveling mean for investor trust and the due diligence processes within the AI sector?

In today’s AI recap:

• Builder.ai's $1.5B collapse amid deception claims

• Meta’s major nuclear energy deal for AI

• Google pauses ‘Ask Photos’ AI rollout

• Spotting AI fakes gets more challenging

AI Startup's $1.5B Implosion

The Recap: Builder.ai, once valued at $1.5 billion and backed by Microsoft, has declared bankruptcy after allegations surfaced that it primarily used human developers to fake its AI capabilities, not advanced AI as claimed.

Unpacked:

• A key lender, Viola Credit, withdrawing $37 million from Builder.ai's accounts crippled its operations across five nations, as Bloomberg reported.

• The company allegedly lacked true AI, instead using a team of Indian developers to manually write code, a deception reportedly sustained for eight years before being exposed in a LinkedIn post.

• Builder.ai had amassed over $450 million in total funding from prominent investors including Microsoft, the World Bank's IFC, and SoftBank's DeepCore incubator, all drawn to its AI-driven app-building promise.

Bottom line: This situation underscores the critical need for thorough due diligence in the booming AI space. It also serves as a stark reminder that transparency and genuine technological capabilities must underpin AI marketing claims to maintain trust.

Meta Bets Big on Nuclear for AI

The Recap: Meta is making a massive 20-year commitment to nuclear energy, partnering with Constellation Energy (as detailed in their official announcement to power its AI ambitions. This move underscores the enormous and growing electricity demands of large-scale artificial intelligence infrastructure.

Unpacked:

• The agreement centers on Constellation's 1.1-gigawatt Clinton Clean Energy Center, a key facility in Illinois, ensuring its operation through 2047 by supporting relicensing, preserving over 1,100 jobs, and boosting output by 30 megawatts.

• Meta joins a growing roster of tech titans, including Microsoft, Amazon, and Google, who are increasingly turning to nuclear to provide the reliable, carbon-free electricity needed for their energy-hungry data centers.

• This strategic shift highlights the immense energy footprint of AI development and the urgent need for scalable, clean energy solutions to power the AI-driven future sustainably, with projections showing data center energy demand soaring globally.

Bottom line: This landmark deal signals a crucial intersection of AI's exponential growth and the pressing need for dependable, large-scale clean energy sources. As AI continues to reshape industries, securing such power supplies is becoming fundamental for technological advancement and meeting ambitious climate goals.

AI Training

Collect Emails from Any Site with AI (free n8n template)

The Recap: In this video, we use n8n and Firecrawl to build an AI-based email scraping workflow that can grab relevant email addresses from any website. These emails can be used in your cold email lead generation campaigns!

P.S We also launched a free community for AI Builders looking to master the art and science of building AI Automations — Come join us!

Google Taps Brakes on AI Photo Search

The Recap: Google has temporarily halted the rollout of its Gemini-powered "Ask Photos" feature, an AI tool designed for natural language search within Google Photos. The pause addresses concerns about latency, quality, and overall user experience before a wider release.

Unpacked:

• "Ask Photos" aims to let users query their photo libraries with conversational questions, leveraging Gemini's multimodal capabilities to understand image content and context.

• A Google Photos product manager, Jamie Aspinall, confirmed the rollout pause on X, stating the feature "isn't where it needs to be" and an improved version is expected in about two weeks.

• This isn't Google's first AI pullback; the company previously adjusted "AI Overview" in Search and its Gemini image generator after both produced inaccurate or problematic results.

Bottom line: This move highlights the significant hurdles tech giants face in deploying advanced AI features reliably and at scale. Ensuring accuracy and a smooth user experience remains paramount as the AI race continues.

Where AI Experts Share Their Best Work

Join our Free AI Automation Community

Join our FREE community AI Automation Mastery — where entrepreneurs, AI builders, and AI agency owners share templates, solve problems together, and learn from each other's wins (and mistakes).

What makes our community different:

• Real peer support from people building actual AI businesses

• Complete access to download our automation library of battle-tested n8n templates

• Collaborate and problem-solve with AI experts when you get stuck

Dive into our course materials, collaborate with experienced builders, and turn automation challenges into shared wins. Join here (completely free).

Spotting AI Fakes Gets Harder

The Recap: The latest AI video tools, like Google's Veo 3, now create incredibly realistic footage, making it much harder to spot fakes and raising alarms about the potential spread of misinformation.

Unpacked:

• Google's Veo 3 can generate hyper-realistic videos with natural lighting and movement from simple text prompts, often good enough to fool casual viewers on social media.

• To combat potential misuse, Google embeds SynthID watermarks in its AI-generated content and recently launched a SynthID Detector portal for users to verify images, audio, video, and text.

• Beyond specific tools, Google aims to grow a broader ecosystem for content authenticity, making its SynthID technology available for other developers and partnering with content verification platforms.

Bottom line: As AI-generated content becomes indistinguishable from reality, developing both advanced detection tools and critical media literacy is essential. This evolving landscape underscores the ongoing need for transparency and user vigilance in the digital age.

The Shortlist

LawZero detailed its mission to advance safe-by-design AI through its "Scientist AI" approach, a non-agentic system for oversight and scientific discovery, launched under the leadership of AI pioneer Yoshua Bengio.

Anthropic launched "Claude Explains," a new blog where its AI model Claude generates most of the content on technical topics, with human subject matter experts and editorial teams reviewing and enhancing the drafts.

TSMC Chairman C.C. Wei expressed confidence that the company will remain the primary manufacturer for AI chips, irrespective of design winners, citing deep collaboration with chip designers and their cloud service provider customers.

COAR reported that its survey found over 90% of open access repositories are facing significant operational issues and service outages due to aggressive content scraping by AI bots for training data.

Signing off,

David, Lucas, Mitchell — The Recap editorial team