# AI Skills'2026 Virtual Conf Source: https://www.youtube.com/watch?v=89dfqzEmxiI Recap page: https://rapidrecap.app/video/89dfqzEmxiI Generated: 2026-01-28T10:37:05.704+00:00 --- ## Quick Overview The AI Skills 2026 Virtual Conference, hosted by Max and co-hosted by God of Prompt, Pabo, and Akash Gupta, focused heavily on pragmatic AI adoption, moving beyond prompt engineering to agentic orchestration, objective design, and context engineering, while highlighting skyrocketing performance expectations across all job functions due to AI capabilities. **Key Points:** - Martin advocates for a 'platform over point solutions' framework for new businesses, but suggests an 80/20 rule for existing enterprises sticking to their current hyperscaler ecosystem like Azure for security and compliance. - Virginia suggests looking to top-tier players like Microsoft or Google for reliable AI features because when big companies invest capital in integrating concepts like grounding and RAG solutions, it signals longevity over chasing fads. - Dimma stresses that production-ready AI products require a clear evaluation framework and metrics to avoid launching expensive experiments, noting that evaluating stochastic GenAI outputs is significantly harder than with classical technology. - Martin asserts that agentic orchestration will be crucial by 2026, stating that the coordination layer—checking on and managing AI agents—becomes the leader's primary job. - Context engineering involves providing models with environmental metadata and descriptions beyond the initial prompt, essentially creating a controlled environment for the model to operate within, as described by Dimma. - The panel agreed that legacy habits like overthinking and being precious over perfect code should be killed; the new macro skills involve rapidly prototyping, architecture diagramming, and defining objective design metrics. - Dimma demonstrated using his local system (Corsur) with custom agents to simulate stakeholder negotiations and draft executive-style communications based on pre-loaded context and style guides. **Context:** The video captures the opening and a key panel discussion from the 'AI Skills 2026 Virtual Conf,' organized by entrepreneur Max, who has been facilitating virtual events for investors and startups. The conference features co-hosts Robert (God of Prompt), Pabo (experimenting with AI agents/no-code), and Akash Gupta (host of the AI product managers podcast). The initial panel, moderated by Liam Dupson of Encounter AI, focused on navigating AI tool noise and identifying essential skills for 2026, featuring speakers Martin (Casper Studios/ex-Microsoft Azure AI), Virginia Roberts (Cloud/AI Trainer), and Dimma Barovski (Science Director at Deliveroo/DoorDash). ## Detailed Analysis The conference immediately established a pragmatic tone, emphasizing real-world application over hype. Martin, speaking from an enterprise GTM perspective, advised prioritizing platform stacks over niche tools unless a clear transformation strategy demands otherwise, acknowledging that hyperscalers introduce product launch delays compared to agile startups. Virginia reinforced looking at large-scale adoption by hyperscalers as validation that a technology, like RAG solutions, has the necessary capital backing to be worth the investment. Dimma strongly cautioned against launching AI projects without defined evaluation metrics, citing millions wasted on unmeasured experiments, especially with unpredictable GenAI outputs. The discussion pivoted to future skills, confirming that prompting is now a commodity, making agentic orchestration a critical skill where leaders manage AI interns. Virginia and Dimma explored objective design—teaching models what success looks like via metrics—and context engineering, where humans codify environmental metadata for agents, which Dimma believes may be humanity's last crucial task before full automation. The lightning round concluded that the immediate focus must be on building prototypes, killing overthinking and preciousness about code, and developing architecture diagramming skills. Dimma later demonstrated using his local agent network to prepare for high-stakes meetings by simulating stakeholder negotiations and drafting communications in specific corporate styles, illustrating how advanced agents are already shaping managerial functions and how performance expectations are skyrocketing, with companies like Uber now requiring candidates to v-p code their energy source proposals. ### Conference Overview - Max hosts the 18th virtual conference, the second focused on AI, welcoming co-hosts Robert, Pabo, and Akash Gupta - Over 1,600 people promoted the event on social media. ### Enterprise AI Adoption Strategy - Martin advises platform-first for startups but 80/20 rule for existing enterprises on hyperscaler stacks - Virginia notes that hyperscaler integration validates that new features like RAG have sufficient capital investment. ### Evaluating AI Success - Dimma states that the key difference between production-ready AI and expensive experiments lies in having an evaluation framework from the start, highlighting the difficulty in measuring stochastic GenAI outputs. ### Future AI Skills - Panelists agree prompting is commoditized; Martin emphasizes agentic orchestration as the key managerial coordination layer for 2026 - Virginia links objective design (teaching models success metrics) to evaluation metrics. ### Context Engineering Defined - Dimma defines context engineering as providing models with environmental metadata and descriptions, moving from narrow tasks to exploration within created limitations. ### Legacy Habits to Kill - Martin suggests killing the 'build, buy, or partner' sequence in favor of immediately building a prototype - Virginia suggests killing being 'precious over code' - Dimma advises killing overthinking and prioritizing building and shipping. ### New Macro Skills - Virginia recommends novice architecture diagramming for everyone to map workflows and spot security/compliance holes - Dimma advocates for building taste by shipping to understand what 'good' looks like. ### Dimma's Agent Demonstration - Dimma showcased agents preparing him for executive meetings by simulating stakeholder negotiations based on uploaded context and drafting communications in specific Slack/executive styles using local processing.