Agents at Work

AMD IS COMING — The War Nobody Was Watching

On Friday April 24 2026, AMD jumped 13.9% to an all-time high — on a day with zero AMD-specific news. CNBC's headline buried the lede: "AMD shares soar 12% on no company news." The catalyst was Intel's blowout Q1 print and a single-sentence DA Davidson upgrade: "the CPU is reinserting itself as an indispensable foundation of the AI era." That sentence is the entire thesis of this episode. Hank walks through the AMD story most coverage gets wrong: — How AMD took 41% of Intel's server CPU market while everyone was watching GPUs — Why chiplet manufacturing put AMD a generation and a half ahead of Intel — The HBM capacity moat (288 GB MI355X vs 192 GB B200) and what it means for inference economics — ROCm vs CUDA in 2026 — and why the inference gap has effectively closed — The OpenAI 6 GW and Meta 6 GW deals (and the ~20% of AMD equity quietly issued in performance warrants) — Three specific predictions for 2027–2030 — The honest bear case: CoWoS allocation, ROCm training gap, Lisa Su succession risk The audience is business owners and operators, not chip designers. Plain-spoken, anti-hype, defensible. The "little brother" framing is over. Read the full written analysis at https://prentusai.com/news/amd-is-coming.html Watch the 25-min video version at https://www.youtube.com/@Prentusai Independent analysis from Prentus AI.

State of AI, Part 1: Why 95% of AI Pilots Fail

An honest look at where AI actually stands in April 2026, and why the MIT Project NANDA study reported that 95% of enterprise AI pilots produce no measurable return. Frontier capability has moved fast: Claude Opus 4.7, GPT-5.4 with native computer use, a Chinese open-weight tier within shouting distance of the US labs, and METR time-horizon numbers that put top models at 3-5 hours of effective autonomous work. But the binding constraint on AI for businesses right now is not capability. It is reliability. And reliability is moving far slower than the capability curve suggests. Part 1 of a two-part series. In this episode: * The real state of the frontier in April 2026 (Opus 4.7, GPT-5.4, Gemini 3.1 Pro, Chinese open weights) * Why METR's "time horizon" metric matters more than one-shot benchmarks * The three forces driving current capability gains (compute scaling, post-training, interpretability) * The limits that do not show up in keynote demos: hallucination rates, long-horizon reliability, the gap between benchmark and real work * The MIT Project NANDA study everyone quotes, and what it actually found * The circular-capital dynamic: $700 billion in 2026 hyperscaler capex and the money flowing between AI suppliers * Why the next binding constraint on AI infrastructure is electrical power, not chips Full written analysis: https://prentusai.com/news/the-state-of-ai-april-2026.html [https://prentusai.com/news/the-state-of-ai-april-2026.html] Part 2 covers geopolitics, the labor picture, the next 12-24 months, and a practical roadmap for a business owner deciding what to do next. Agents at Work is the podcast of PrentusAI, an independent learning hub covering OpenClaw, NemoClaw, and AI agent technology in plain English for business owners. Subscribe for weekly deep dives. Web: https://prentusai.com [https://prentusai.com/] Newsletter: https://prentusai.com [https://prentusai.com/] Full article archive: https://prentusai.com/news/ [https://prentusai.com/news/]

EP 3: The Decision — EU AI Act, Costs, and What to Do Now

The EU AI Act hits August 2 with 7% revenue penalties. Self-hosting isn't always cheaper. Here's the honest cost comparison and six things every business owner should do right now. Part 3 of 3. | prentusai.com

EP 2: The Open-Source Counter — OpenClaw's Rise and Security Crisis

penClaw hit 346K GitHub stars in 5 months and 3.2M users. But 42,665 instances are exposed, 36% of ClawHub skills contain prompt injection, and Kaspersky called it unsafe. Enter NemoClaw. Part 2 of 3. | prentusai.com

EP 1: The Announcement — Claude Managed Agents Explained - Description: Anthropic just launched Claude Managed Agents. We bre

Anthropic just launched Claude Managed Agents. We break down what it actually is, what it costs ($0.08/session-hour + tokens), and which major companies are already using it, including Snowflake's $200M bet. Part 1 of 3. | prentusai.com