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 29, 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 29, 2026. You're listening to the DX Today AIDGay brief. Today, Softbank borrows$40 billion to double down on OpenAI. Apple poaches a Google executive to fix its AI marketing problem. And SK Heinex files for a Blockbuster US listing to feed the AI memory boom. Let's get into it.
SPEAKER_04Softbank Group has secured a$40 billion bridge loan, the largest dollar-denominated loan of its kind, to fund its$30 billion additional investment in OpenAI. The one-year unsecured facility was arranged by JP Morgan and Goldman Sachs. Softbank says it will partially repay the loan through the sale of existing assets. The investment was first announced in February as part of OpenAI's broader$110 billion funding round, which valued the company at roughly$840 billion. The deal significantly increases Softbank's debt profile, but cements its position as one of the largest backers of AI infrastructure in the world. CEO Masayoshi Sun has described artificial intelligence as the defining investment opportunity of the century.
SPEAKER_03Apple recruits from a rival.
SPEAKER_00Apple has hired Lillian Rinken, a former Google vice president, GGay lead product marketing for artificial intelligence. Rinken spent nearly a decade at Google, most recently overseeing Google Shopping, and will now report directly GG Greg Joswiak. Her arrival comes as Apple G Gay prepares a long-delayed overhaul of its Siri assistant, powered by Google's GDK models under a multi-year collaboration announced earlier this year. The hire signals that Apple recognizes it needs fresh leadership to communicate its AI story, GDK, who have grown skeptical of the company's ability to compete in the space. Apple's AI features have lagged behind rivals, and the Gemini-powered Siri upgrade, originally targeted for earlier this year, has GDK pushed to a future iOS release.
SPEAKER_03A chip giant eyes wool street. S.K.
SPEAKER_02Heinex, the South Korean memory chipmaker that supplies high bandwidth memory, critical to AI systems from companies like NVIDIA, has confidentially filed to list in the United States. The offering, targeting the second half of 2026, could raise an estimated$10 to$14 billion. The move is driven by a persistent valuation gap with global peers like Micron and a need to fund the massive capital expenditures required to meet surging AI demand. At its annual shareholder meeting, CEO No Jung Quak said the company is targeting approximately$75 billion in net cash to support long-term investments. SK Heinox has also signed a$7.9 billion deal to acquire advanced lithography equipment from ASML to boost memory production for AI workloads. OpenAI's next model takes shape.
SPEAKER_05OpenAI has completed training on its next major AI model, internally codenamed SPUD. CEO Sam Altman told employees the model is very strong and could be released within weeks, describing it as something that can really accelerate the economy. The freed-up compute from shutting down Sora is being redirected to support the new model's development and deployment. Details remain scarce. The parameter count, architecture, and whether it is a reasoning or multimodal model have not been disclosed, nor has its commercial name. Some analysts speculate it could represent GPT 5.5 or even GPT-6. The company is simultaneously consolidating ChatGPT, Codex, and its Atlas web browser into a unified super app that the new model will power.
SPEAKER_03AI coding gets faster feedback loops.
SPEAKER_01Cursor, the AI-powered code editor, has revealed that it uses a technique called real-time reinforcement learning to improve its composer coding model in production. The approach involves serving model checkpoints directly to users, observing their responses, and aggregating those signals as training rewards. The result is a remarkably fast iteration cycle. Cursor says it can ship an improved version of Composer as often as every five hours. The technique builds on the company's earlier disclosure that Composer 2 was trained on top of Moonshot AI's open source Kimmy model, with about three-quarters of the compute coming from Cursor's own additional training. The approach reflects a growing trend where AI companies combine open source foundations with proprietary reinforcement learning to create differentiated products at lower cost.
SPEAKER_04Perplexity has launched Perplexity Health, a suite of data connectors that links a user's electronic health records, wearable device data, and lab results into its AI search platform. The product connects to Apple Health, Fitbit, UltraHuman, and Withings with electronic health records pulled through Be Well Connected Health, which covers more than 2.4 million providers and 350 health plans. Users can ask questions grounded in their actual medical history rather than generic information. Perplexity Computer, the company's agent platform, can generate personalized outputs, including pre-appointment summaries, nutrition plans, and training protocols. The company has launched a health advisory board and emphasizes that the product is educational, not diagnostic. It is rolling out to pro and max subscribers in the United States.
SPEAKER_03An AI engineer finds its footing.
SPEAKER_00Cognition's Devon, the autonomous AI software engineer, is gaining traction with major enterprise deployments. At Brazilian FinTech giant New Bank, Devon completed code migration tasks in weeks that would have taken human engineers months, delivering a 12 times improvement in engineering hours and more than 20 times cost savings. The system operates independently in a sandboxed cloud environment, planning, coding, testing, and debugging, with a human kept in the loop only to manage the project and approve changes. Cognition says Devon improves with experience, avoiding rabbit holes more reliably and finding faster solutions as it encounters more examples. At$500 per month, it remains a fraction of the cost of a human engineer, though the company is reportedly exploring lower pricing tiers for smaller teams.
SPEAKER_03Claude learns to dream.
SPEAKER_02The system periodically scans Claude's stored notes, identifies what is still relevant, merges overlapping entries, removes outdated information, and standardizes details like replacing relative timestamps with fixed dates. AutoDream builds on Anthropic's auto memory feature, launched in February, which lets Claude save its own notes across coding sessions. Early community analysis suggests AutoDream runs infrequently, triggering only after certain thresholds of usage and elapsed time. It does not modify the underlying code base, operating only on memory files. The feature addresses a fundamental challenge for long-running AI agents, keeping accumulated context organized and useful without human intervention. A digital twin of the human brain.
SPEAKER_05Meta has released Tribe V2, a foundation model trained to predict how the human brain responds to virtually any sight or sound. Building on architecture that won the Algonauts 2025 competition, the model draws on more than 500 hours of functional MRI recordings from over 700 people to create what Meta calls a digital twin of neural activity. Tribe V2 enables zero-shot predictions for new subjects, languages, and tasks, consistently outperforming standard modeling approaches. Meta is releasing the model, code base, research paper, and an interactive demo. The company says the technology could help researchers advance neuroscience, apply brain insights to build better AI systems, and use computational simulation to accelerate breakthroughs in diagnosing and treating neurological diseases.
SPEAKER_03Humanoid robots ship for real.
SPEAKER_01Boston Dynamics has begun shipping its production-ready Atlas Humanoid Robot, and the initial run is already sold out. The fully electric fifth-generation robot operates on custom direct drive actuators and uses a machine learning vision model to detect objects and generate motions in real time. Boston Dynamics partnered with Google Deep Mind to integrate foundation models into the robot, giving it a general intelligence layer on top of its physical capabilities. Each unit is priced at approximately$420,000, deliberately below the cost of employing two manufacturing workers for two years. Parent company Hyundai has announced a$26 billion investment in U.S. operations, including a factory capable of producing 30,000 Atlas units per year by 2028, when the per unit price is expected to drop to around$200,000.
SPEAKER_03AI transforms a centuries-old company.
SPEAKER_04OpenAI is deploying its enterprise platform at Stadler, a 230-year-old Swiss rail vehicle manufacturer. The partnership is being cited as a case study in how AI is moving beyond technology companies into established industrial sectors. Stadler is using OpenAI's tools to accelerate productivity and transform traditional knowledge work, including documentation, engineering workflows, and operational processes. The deployment reflects OpenAI's broader push under Fijicmo to refocus the company on enterprise customers ahead of a potential IPO. The Stadler implementation shows that the enterprise AI market is expanding well beyond Silicon Valley, with manufacturers, logistics firms, and industrial companies now adopting agent-based tools to streamline complex operations that have traditionally relied on manual processes and institutional knowledge.
SPEAKER_03And finally, the valuation question.
SPEAKER_00A growing chorus of analysts is warning that the valuation gap between closed source and open source AI companies may be narrowing faster than investors realize. Open source models are reaching performance parity with Frontier Lab offerings on key benchmarks, raising questions about whether companies like OpenAI and Anthropic can sustain the premium pricing that justifies their combined valuations of well over a trillion dollars. The key metric, according to one widely discussed analysis, is not the capability spread between models, but the monetizable spread, the subset of that gap that enterprises will actually pay a premium for. As open weight alternatives from Meta, Mistral, Alibaba, and others continue to improve, the monetizable spread is declining faster than raw capability differences, potentially compressing margins across the industry.
SPEAKER_03That's your briefing for Sunday, March 29th. For DX Today, stay curious.