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 - Sunday, March 22, 2026
Use Left/Right to seek, Home/End to jump to start or end. Hold shift to jump forward or backward.
It's Sunday, March 22, 2026. You're listening to the DX Today AI Daily Brief. This morning, OpenAI plans to nearly double its workforce to 8,000. The Pentagon formalizes Palantir's Maven AI as a core military system. And AI startups have captured a record 41% of all venture capital. Plus, Meta brings its MANIS AI agent to the desktop. A new open source architecture challenges the Transformer, and two-thirds of Fortune 500 CEOs are freezing hiring while betting billions on AI. Let's get started.
SPEAKER_01OpenAI plans to nearly double its workforce from 4,500 to 8,000 employees by the end of 2026, according to the Financial Times. The majority of new hires will be in product development, engineering, research, and sales. The company is also recruiting specialists it calls technical ambassadors, whose role is to help businesses optimize their use of OpenAI's tools. The hiring push comes as OpenAI sharpens its focus on enterprise customers and prepares for a potential IPO as early as the fourth quarter of this year. It also stands in stark contrast to the broader tech industry, where nearly 40,000 jobs have been cut in the first quarter alone. OpenAI's annualized revenue stands at roughly$25 billion.
SPEAKER_03A landmark move at the Pentagon.
SPEAKER_00The Pentagon will formally adopt Palantir's Maven Artificial Intelligence System as an official program of record, according to a memo obtained by Reuters. Deputy Secretary of Defense Steve Feinberg communicated the decision to Pentagon leadership on Friday. The designation locks in long-term use of Palantir's weapons targeting and intelligence analysis technology across the U.S. military. Maven, which began as a controversial AI project inside Google in 2018, has evolved under Palantir into one of the most widely deployed military AI platforms. The move to program of record status means Maven will receive dedicated funding and institutional support, expected to take full effect before the fiscal year ends in September. Hallantier shares Rose on the news.
SPEAKER_03Continued fallout from Friday's policy release. Legal analysts are weighing in on the White House National AI policy framework released Friday.
SPEAKER_02The document, organized around seven pillars, calls on Congress to preempt state AI laws and establish a single federal regulatory standard. It recommends against creating any new federal rulemaking body, instead, directing existing agencies to handle sector-specific AI oversight. The framework also urges Congress to codify the Trump administration's ratepayer protection pledge, which requires AI companies to supply or pay for electricity consumed by their data centers. Legal experts at the National Law Review note that while the framework aligns with prior executive orders, its implementation through Congress remains uncertain, particularly given the absence of bipartisan support for such sweeping preemption.
SPEAKER_03Now to AI agents on the desktop.
SPEAKER_05Meta's AI agent MANIS has launched a desktop application for Mac and Windows, bringing its autonomous task execution capabilities directly to personal devices. The app, which starts at$20 a month, allows users to delegate complex workflows across files, applications, and the web to an AI agent that operates independently. MANIS was originally an independent AI startup that Meta acquired earlier this year, folding the team into its Superintelligence Labs division. The desktop launch places Meta in direct competition with Anthropic's co-work, OpenAI's Codex, and a growing ecosystem of agent-based productivity tools. A free tier offers limited access, but the full capabilities require a paid subscription.
SPEAKER_03The venture capital numbers are in.
SPEAKER_04AI startups captured a record 41% of the$128 billion in venture funding tracked by Carta last year, according to data reported by TechCrunch. That translates to more than$52 billion flowing into AI companies in a single year. Just three years ago, AI accounted for roughly 15% of venture funding. The sector has nearly tripled its share, pulling capital away from consumer apps, fintech, and crypto. Carta data also shows that funds raised in 2023 and 2024, after the launch of ChatGPT, have posted the highest internal rate of return compared with declining returns from funds raised between 2017 and 2020. The numbers validate the enormous bet the venture industry has placed on artificial intelligence.
SPEAKER_01Mamba 3, a new open source state-space model, has arrived with claims of surpassing transformer architecture on speed, cost, and inference efficiency. Published at ICLR 2026 by the original Mamba authors, including Albert Gu, the model introduces three core improvements. A more expressive recurrence derived from state-space discretization, a complex valued state update rule that enables richer state tracking, and a multi-input, multi-output variant that expands internal processing capacity. Unlike transformers, which store past information in a growing key-value cache, Mamba 3 uses a fixed-size internal state that does not scale with sequence length. At the 1.5 billion parameter scale, the model achieved the fastest combined prefill and decode latency across tested sequence lengths, opening a potential path to cheaper, faster AI inference.
SPEAKER_03Enterprise AI browsing gets a security layer.
SPEAKER_00Perplexity has launched Comet Enterprise, a team-ready version of its AI-powered browser, with a significant security partnership with CrowdStrike. The integration embeds CrowdStrike's Falcon platform directly into the browser, providing real-time threat detection, data governance, and protection against unauthorized information sharing during AI-driven workflows. Enterprise administrators can control where and how Comet operates, enabling permissions across the browser or restricting it to approved domains. The platform also connects to 20 frontier language models and integrates with enterprise systems like Snowflake and Salesforce through app connectors. Perplexity says the enterprise version is designed for everything from customer meeting preparation to contract analysis and financial modeling.
SPEAKER_02The new offering pairs NVIDIA's training optimized GPUs with GROKS chips, which are designed specifically for fast, cost-effective AI inference. The partnership addresses a growing competitive vulnerability for NVIDIA, as rivals like Google's TPUs and Cerebrus have been winning inference workloads from major clients, including OpenAI and Meta. Analysts say the Grok integration, which follows a$20 billion licensing deal signed in December, could help NVIDIA reclaim ground in the rapidly expanding inference market. CEO Jensen Huang told attendees that the turning point for inference has arrived as AI systems increasingly perform productive, revenue-generating tasks.
SPEAKER_05The CMA had proposed changes requiring Google to let publishers exclude their content from AI overviews and from broader AI model training. While Google disputes claims that AI search features harm publishers, the company says it is building the opt-out mechanism. Details and timelines remain unclear. The move comes as publishers worldwide grow increasingly concerned about AI systems consuming their content to generate answers that reduce traditional website traffic. Google must also address worries that opting out of AI overviews could negatively impact a site's ranking in standard search results.
SPEAKER_03AI code review goes multi-agent.
SPEAKER_04Instead of relying on a single reviewer, the system assigns agents to scan code changes, check the surrounding code base for context, verify potential bugs, and rank issues by severity before posting a consolidated review. Internal data shows that for large pull requests exceeding 1,000 lines, 84% surface issues, averaging seven and a half findings per request. Engineers report less than 1% of findings are marked incorrect. In one case, a routine-looking one-line production change was flagged as critical because it would have broken authentication. Anthropic has been using the system internally for months.
SPEAKER_01According to a fortune analysis of corporate workforce strategies, the labor market has softened enough for companies to shrink headcount without the stigma of mass layoffs. And by early 2026, that retrenchment had hardened into institutional policy. Nearly 40,000 tech jobs have been cut in the first quarter across roughly 60 companies. Fortune reports that companies including Meta, Amazon, Salesforce, and Workday have announced reductions, with many explicitly citing AI-driven efficiencies. The pattern represents a fundamental shift in how large organizations are architecting their workforces, replacing roles with AI systems rather than simply automating existing workflows.
SPEAKER_00And finally, a new model from the French lab, Mistral AI, has released Mistral Small 4, a unified, multimodal model built on a mixture of experts architecture with 119 billion parameters. The model integrates the capabilities of three previously separate Mistral products, Magistral for reasoning, Pixel for vision, and Devstrel for coding, into a single system that supports both text and image inputs. It features configurable reasoning effort, allowing developers to balance accuracy and speed depending on the task. Mistrel Small 4 is fully open source and available on platforms including VLLM, Llama.cpp, and Hugging Face Transformers. The release reflects a broader industry trend of consolidating specialized capabilities into smaller, more efficient models rather than building ever larger, monolithic systems.
SPEAKER_03That's your DX Today AI Daily Brief for Sunday, March 22nd, 2026. Twelve stories across hiring, defense, policy, venture capital, and the race to build faster, cheaper AI. For DX Today, stay curious.