DX Today | No-Hype Podcast & News About AI & DX
The DX Today Podcast: Real Insights About AI and Digital Transformation
Tired of AI hype and transformation snake oil? This isn't another sales pitch disguised as expertise. Join a 30+ year tech veteran and Chief AI Officer who's built $1.2 billion in real solutions—and has the battle scars to prove it.
No vendor agenda. No sponsored content. Just unfiltered insights about what actually works in AI and digital transformation, what spectacularly fails, and why most "expert" advice misses the mark.
If you're looking for honest perspectives from someone who's been in the trenches since before "digital transformation" was a buzzword, you've found your show. Real problems, real solutions, real talk.
For executives, practitioners, and anyone who wants the truth about technology without the sales pitch.
DX Today | No-Hype Podcast & News About AI & DX
DX Today AI Daily Brief - Friday, March 27, 2026
Use Left/Right to seek, Home/End to jump to start or end. Hold shift to jump forward or backward.
It's Friday, March 27, 2026. You're listening to the DX Today AI Daily Brief. Today, a data leak exposes Anthropic's most powerful AI model yet. Google launches a real-time audio model across 200 countries, and a federal judge blocks the Pentagon from blacklisting an AI company over its safety stance. Let's get into it.
SPEAKER_06Anthropic is scrambling after an accidental data leak revealed the existence of Claude Mythos, what the company calls the most capable AI model it has ever built. Internal draft documents surfaced in a publicly accessible data store due to a content management system error, exposing details of the model in a new classification tier called Capibara, which sits above Anthropic's current top-tier OPUS models. According to the Leak Drafts, Mythos dramatically outperforms Claude Opus 4.6 in coding, academic reasoning, and cybersecurity tasks. Anthropic confirmed the model exists and is being tested with a small group of early access customers. The company flags serious concerns that the model's advanced cyber capabilities could be exploited by hackers to find and attack software vulnerabilities at scale, outpacing current defensive tools.
SPEAKER_02Google rethinks Voice AI.
SPEAKER_00Google has released Gemini 3.1 Flash Live, a specialized model designed to make AI voice conversations feel genuinely human. The model powers both Gemini Live and Search Live, offering lower latency and improved tonal understanding. It can detect subtle acoustic cues like pitch and pace, adjusting its tone and response length to match the moment. In benchmark testing, the model scored 90.8% on complex funk bench audio, which measures how well an AI follows multi-step instructions in real time. Google is also using the model to expand Search Live globally to more than 200 countries, supporting over 90 languages. The release includes synth ID watermarking technology that embeds imperceptible markers in AI-generated audio to help identify synthetic speech.
SPEAKER_02Trump's AI czar calls it quits.
SPEAKER_01David Sachs, the venture capitalist who served as President Trump's artificial intelligence and cryptocurrency advisor, has completed his 130-day stint as a special government employee. Sachs confirmed in an interview with Bloomberg that he is transitioning to co-chair the President's Council of Advisors on Science and Technology, known as PCAST, alongside senior White House technology advisor Michael Kratios. Sachs said the new role will allow him to advise on a broader range of technology issues beyond AI, including advanced semiconductors, quantum computing, and nuclear power. His departure comes just one week after the White House released its national AI legislative framework. A source familiar with the matter told Axios that the White House does not plan to appoint a new AI czar, leaving the position vacant. Defense AI draws massive funding.
SPEAKER_05SHIELD AI, the defense technology startup behind the autonomous drone pilot system HiveMind, has raised$2 billion in a Series G round at a$12.7 billion valuation, more than double its valuation from a year ago. The round was led by Advent International with significant participation from J.P. Morgan Chase's Strategic Investment Group and existing backers, Snowpoint Ventures and Riot Ventures. Shield AI's software enables military drones and aircraft to operate autonomously in GPS-denied environments without requiring remote human pilots. The massive raise underscores surging investor appetite for defense AI as autonomous systems play an increasingly central role in modern warfare. The company plans to use the capital to scale hive mine deployments and expand its simulation capabilities through a recent acquisition.
SPEAKER_02Anthropic wins in court.
SPEAKER_03A federal judge has temporarily blocked the Pentagon from designating Anthropic as a supply chain risk, handing the AI company a significant legal victory. U.S. District Judge Rita Lin ruled that the Defense Department's actions appeared aimed at punishing Anthropic rather than protecting national security. The blacklisting, the first ever applied to a U.S. company under this designation, was imposed after Anthropic refused to allow the military to use its clawed AI software for surveillance or autonomous weapons. Judge Lin also blocked enforcement of a Trump administration's social media directive ordering all federal agencies to stop using anthropic products. The ruling takes effect in seven days to allow time for a potential appeal. Anthropic said it was pleased the court agreed it is likely to succeed on the merits of its case.
SPEAKER_02A new voice for open source.
SPEAKER_06Enterprise AI company Cohere has launched Transcribe, its first open source automatic speech recognition model. At 2 billion parameters, the model is designed to run on consumer great hardware and currently supports 14 languages, including English, Chinese, Arabic, and Japanese. Cohere says Transcribe tops the Hugging Face OpenASR leaderboard with an average word error rate of 5.42, outperforming models from Zoom, IBM, Eleven Labs, and Alibaba's Quen team. The model can process 525 minutes of audio in a single minute, making it one of the fastest in its class. Cohere plans to integrate Transcribe into its North platform for enterprise agent orchestration and is offering free API access. The launch comes as demand for AI-powered transcription and dictation tools continues to accelerate across business environments.
SPEAKER_02Mistral enters the speech race.
SPEAKER_00French AI lab Mistral has released Voxtrel TTS, an open-weight text-to-speech model that the company says rivals or exceeds commercial alternatives at a fraction of the size. At just 4 billion parameters, the model is small enough to run on a laptop, yet generates emotionally expressive speech in nine languages with a latency of just 70 milliseconds. Voxtrel uses a novel architecture combining a transformer decoder, a flow-matching acoustic transformer, and an in-house neural audio codec. It also supports zero-shot cross-lingual voice adaptation, meaning it can generate speech in one language using a voice sample from another, preserving the original accent. The model is available as open weights, positioning Mistral as a serious competitor to proprietary services from 11 Labs and others in the rapidly growing voice agent market.
SPEAKER_02New chips to cut AI energy costs. Normal Computing has raised$50 million in a round led by Samsung's Catalyst Fund, bringing its total funding to more than 85 million. The startup is tackling two converging problems in the semiconductor industry: the growing complexity of chip design and the surging energy demands of AI infrastructure. Normal Computing builds AI software that helps chip companies design silicon more efficiently, while also developing its own physics-based computing hardware aimed at dramatically improving performance per watt. The round included participation from Eric Schmidt's First Spark Ventures, Micron Ventures, and the UK's Advanced Research and Invention Agency. Normal is part of a growing wave of startups exploring alternatives to conventional AI hardware, as data center energy consumption continues to climb. The dictionary takes on OpenAI.
SPEAKER_05Encyclopedia Britannica and its Merriam-Webster subsidiary have filed a copyright lawsuit against OpenAI in Manhattan federal court. The publishers allege that OpenAI scraped nearly 100,000 copyrighted articles to train its language models without permission. The complaint also accuses ChatGPT of generating responses that closely reproduce Britannica's content and divert users away from its websites. Britannica further claims that OpenAI violates trademark law when ChatGPT produces hallucinated information and falsely attributes it to the publisher. OpenAI has responded that its models are trained on publicly available data and that the use constitutes fair use because the technology transforms information into new outputs. The case adds to a growing wave of copyright litigation against AI companies from publishers, artists, and content creators.
SPEAKER_02AI Agents Go Enterprise in China.
SPEAKER_03Alibaba has launched Wukong, an enterprise platform designed to orchestrate multiple AI agents performing tasks like document editing, approvals, and research simultaneously. The system integrates with messaging platforms and existing enterprise tools, reflecting a broader industry shift toward agent-based workflows in business environments. Wukong allows companies to deploy and manage fleets of specialized agents that can collaborate on complex processes rather than relying on a single general-purpose chatbot. The launch comes amid intense competition in the Chinese AI market, where companies including Baidu, ByteDance, and Tencent are all racing to build enterprise agent platforms. Alibaba says the system is already being tested by several large Chinese corporations across finance, logistics, and manufacturing sectors.
SPEAKER_02Custom models on demand.
SPEAKER_06Mistral has introduced Forge, a platform that lets enterprises train AI models entirely on their own data rather than fine-tuning pre-built systems. Unlike typical approaches that layer proprietary information onto existing models, Forge enables full customization from the ground up, giving companies greater control over performance, behavior, and regulatory compliance. The platform includes infrastructure, development tools, and embedded engineering support. Early adopters include major industrial and government organizations seeking AI systems that meet strict data sovereignty and security requirements. Mistral says Forge addresses a growing demand from enterprises that need models tailored to highly specialized domains where off-the-shelf solutions fall short. The launch positions Mistrel as a key player in the emerging market for bespoke enterprise AI.
SPEAKER_02And finally, ads may be coming to AI. Google has signaled that it may introduce advertising into its Gemini AI assistant, a move that could reshape the digital advertising landscape. As AI-generated answers increasingly reduce traditional search traffic, integrating ads into conversational interfaces represents a potential new revenue stream for the company. The shift would require entirely new ad formats embedded within AI responses, raising questions about user trust, transparency, and regulatory compliance. Industry observers expect experimentation to begin soon as Google navigates the tension between monetization and user experience. If implemented, the change could set a precedent for how AI platforms across the industry generate revenue, moving beyond subscriptions and API fees to advertising supported models familiar from the search era. That's your briefing for Friday, March 27th. For DX Today, stay curious.