The Snippet

9 US AI startups have raised $100M+ including large language model company Anthropic, AI legal tech company Harvey and synthetic voice startup ElevenLabs Here are the key takeaways: 1/ Anthropic leads the pack with a massive $3.5 billion Series E round that valued the company at $61.5 billion, making it the only US AI startup to secure a round larger than $1 billion so far in 2025. 2/ The healthcare sector is seeing significant AI investment with Abridge valued at $2.75 billion for its clinical conversation transcription technology and Hippocratic AI raising $141 million for its healthcare-focused large language models. 3/ Legal tech is emerging as a hot AI vertical with both Harvey raising $300 million at a $3 billion valuation and Eudia securing $105 million in Series A funding for AI-powered legal solutions. 4/ AI infrastructure companies are attracting substantial capital with Lambda raising $480 million at a $2.5 billion valuation and Together AI securing $305 million for its open source generative AI development infrastructure. 5/ Major tech investors including Nvidia, Andreessen Horowitz, General Catalyst, and Sequoia are heavily backing multiple AI startups across different rounds, signaling continued institutional confidence in the AI sector. For more breakdowns on technology, join our community: https://thesnippet.tech

Big tech ads business have grown 15%+ last year including Meta, Microsoft, and Amazon Here are the key takeaways: 1/ Amazon's ad business reached $17.29 billion in Q4, growing 18% year-over-year, slightly below analyst expectations of $17.4 billion. 2/ Google's parent Alphabet reported 11% growth in Google advertising and 14% growth in YouTube ad revenue, reaching $72.46 billion and $10.47 billion respectively. 3/ Meta leads the pack with 21% year-over-year growth, reporting $48.39 billion in Q4 sales, maintaining its position as the second-largest platform in global digital advertising. 4/ Microsoft's search and news advertising grew an impressive 21% year-over-year, though the company doesn't disclose specific quarterly figures for this segment. 5/ Social media platforms are also seeing strong ad growth, with Pinterest reporting an 18% increase to $1.15 billion and Snap showing 14% growth to $1.56 billion in Q4. For more breakdowns on technology, join our community: https://thesnippet.tech

ChatGPT doubled weekly active users to 400M in last few months and DeepSeek has become the second most visited GenAI tool Here are the other key GenAI trends: 1/ ChatGPT took just 2 months to become the fastest app ever to reach 100M monthly active users and consistently most used GenAI tool 2/ DeepSeek surged to #2 globally in just 10 days, capturing 15% of ChatGPT's mobile user base with slightly higher engagement than Perplexity and Claude users. For more breakdowns on technology, join our community: https://thesnippet.tech 3/ Other popular GenAI tools include CharacterAI, JanitorAI, and QuilBot

Microsoft CEO recently spoke on his belief the need to continuously “refound” Microsoft especially in age of AI and quantum computing Here are the other key takeaways: 1/ He believes "success brings challenges of success" because companies forget what made them successful as new employees think they caused the success rather than recognizing they rode the wave. 2/ While Microsoft had "real daylight" between them and competitors in the 90s, today's tech landscape is more vibrant with "7-8 companies plus anyone can come from anywhere" to create a generational company. 3/ When becoming CEO, Nadella regrounded Microsoft in its mission of "empowering people and organizations" and introduced growth mindset culture to galvanize energy around strategies that were already in place. 4/ On partnerships, Satya views Microsoft fundamentally as both a "platform company and a partner company" seeking long-term win-win constructs, as evidenced by strategic investments in Facebook, OpenAI and others. 5/ Looking ahead, Nadella identifies overlooked opportunities in system architecture supporting AI workloads and distributed model architecture that could leverage computing power across both cloud and local devices. For more breakdowns on technology, join our community: https://thesnippet.tech

A new study shows ChatGPT does not take as much energy as everyone believes (especially compared to one hamburger) Here are the key takeaways: 1/ A recent Epoch AI analysis found ChatGPT consumes around 0.3 watt-hours per query using GPT-4o, which is 10x less than previous estimates of 3 watt-hours. 2/ The energy consumption of ChatGPT is "not a big deal compared to using normal appliances or heating or cooling your home, or driving a car," according to Epoch's data analyst Joshua You. 3/ Despite current efficiency, AI power consumption is expected to increase as models become more advanced, with reasoning models requiring more computing power for their "thinking" process. 4/ By 2030, training a frontier AI model could demand power output equivalent to eight nuclear reactors (8 GW), while AI data centers may need nearly all of California's 2022 power capacity within two years. 5/ Consumers concerned about their AI energy footprint can reduce impact by using ChatGPT sparingly or selecting smaller models like GPT-4o-mini that require less computing power. For more breakdowns on technology, join our community: https://thesnippet.tech

An OpenAI study on AI coding showed surprising findings including that competitor Anthropic Claude was better than their own model and that AI can fix bugs but struggles finding them Here are the key takeaways: 1/ In the SWE-Lancer benchmark, Claude 3.5 Sonnet performed best, earning only $208,050 of the potential $1 million from freelance coding tasks and resolving just 26.2% of individual contributor issues. 2/ OpenAI researchers found that AI models can quickly locate code issues using keyword searches but struggle to understand how problems span multiple components, leading to incomplete or incorrect solutions. 3/ The benchmark tested three frontier models (GPT-4o, o1, and Claude 3.5 Sonnet) on 1,488 real-world freelance software engineering tasks from Upwork, revealing significant reliability gaps for professional deployment. 4/ Models performed better on management tasks requiring reasoning and technical evaluation than on actual coding implementation, suggesting AI might complement rather than replace human engineers. 5/ Despite Sam Altman's claim that models can replace "low-level" engineers, the study concluded that current LLMs lack the ability to properly diagnose root causes of bugs, making them unable to fully replace human software engineers yet. For more breakdowns on technology, join our community: https://thesnippet.tech

7 new tech unicorns were minted in 2025, led by AI companies Here are the companies: 1/ Kikoff, a personal finance platform founded in 2019, hit $1 billion valuation with backing from Lightspeed Venture Partners and NBA star Steph Curry. 2/ Netradyne reached $1.35 billion valuation after raising a $90 million Series D for its computer vision technology, with Point72 Ventures leading the round. 3/ Hippocratic AI, founded just last year in 2023, is already valued at $1.64 billion after a Kleiner Perkins-led $141 million Series B for its healthcare models. 4/ Truveta, a genetic research company founded in 2020, secured a massive $320 million round reaching unicorn status with backing from Microsoft and Regeneron CVCs. 5/ Clay, an AI sales platform established in 2017, hit $1.25 billion valuation after raising a $40 million Series B with investors including Sequoia and First Round. 6/ Mercor, a contract recruiting startup founded in 2022, secured the highest valuation at $2 billion after a $100 million Series B backed by Peter Thiel and Jack Dorsey. 7/ Loft Orbital joined the unicorn club at $1 billion valuation after raising a $170 million Series C for its satellite technology, with Temasek among the investors. For more breakdowns on technology, join our community: https://thesnippet.tech

Google co-founder Sergey Brin in a leaked memo called for employees to be in the office day and work harder at ~60 hours a week to fuel Google to achieve artificial general intelligence (AGI) Here are the key takeaways: 1/ Brin emphasized that "60 hours a week is the sweet spot of productivity" while calling out employees who "put in the bare minimum to get by" as "not only unproductive but also highly demoralizing to everyone else." 2/ The Google co-founder's urgency comes from his belief that "competition has accelerated immensely and the final race to A.G.I. is afoot," suggesting Google is in a critical competitive phase. 3/ Brin believes Google has "all the ingredients to win this race" to artificial general intelligence, but needs to "turbocharge" its efforts to succeed. 4/ He specifically highlighted the need for Google employees to use more of the company's own AI for coding, suggesting this self-improving cycle would lead to AGI. 5/ Brin called on employees working on Gemini to become "the most efficient coders and AI scientists in the world by using our own AI," emphasizing dogfooding their own technology. For more breakdowns on technology, join our community: https://thesnippet.tech

Google Chief Scientist Jeff Dean was recently interviewed, covering topics such as the AI revolution, how 25% of Google code is now AI-generated, future expectations world GDP fueled by AI workers, and more Here are the key takeaways: 1/ Dean envisions massive economic growth from AI, predicting "the world GDP is almost certainly going to go way, way up, two orders of magnitude higher than it is today, due to the fact that we have all of these artificial engineers working on improving things." 2/ Jeff Dean revealed that "25% of the characters that we're checking into our code base these days are generated by our AI-based coding models with kind of human oversight," showcasing AI's growing role in software development. 3/ Jeff Dean explains how Google's early work on language models laid groundwork for today's AI, noting "it felt like neural nets were the right abstraction because it felt like they could solve tiny toy problems that no other approach could solve at the time." 4/ Dean described their vision to have "100 teams around Google" and "people all around the world working to improve languages they care about or particular problems they care about and all collectively work on improving the model." 5/ Dean revealed his approach to innovation through an internal slide deck called "Go, Jeff, Wacky Ideas" which helps "articulate interesting directions you think we should go" without directly ordering teams. For more breakdowns on technology, join our community: https://thesnippet.tech

Perplexity CEO Aravind Srinivas was recently interviewed, covering how he has the confidence to compete against Google, future of web search, and more Here are the key takeaways: 1/ Srinivas said Perplexity's competitive advantage comes from their "obsession about the user and good product taste" while Google faces "a dilemma" as "search kind of exists in service of the ads company." 2/ Srinivas bet on advancing AI with "a very simple heuristic version" that succeeded because "when the models began to be much better instruction following, the dumb approach actually works." 3/ Srinivas believes "there are probably only two problems where you can work on AI and also build product at the same time: one is search and the other is self driving car" because both create data flywheels as improving the product improves the AI. 4/ Srinivas admitted to Perplexity's revenue challenge is that "where do you actually go and fulfill the demand? You go to Google and who gets credit for that monetarily? Google. We got nothing." - indicating future Perplexity features to expand into this space 5/ Srinivas asserted that search leader of tomorrow needs "this amazing orchestration of small models, knowledge graphs, widgets, LLM streaming answers" because "whoever builds that... is going to be the next Google." For more breakdowns on technology, join our community: https://thesnippet.tech