Curt Larson

Noticing an interesting convergence lately. Both Andrej Karpathy and Peter Steinberger are leaning toward CLI as the primary interface for cross-service communication in agent systems instead of MCP. The appeal is obvious: simplicity, composability, and near-universal adoption. I agree that CLI lowers friction and accelerates ecosystem momentum in a way heavier protocols often struggle to match. However, the security surface here deserves more attention. Sandboxing can help, but only to a point, as it would limit what agents can do. When LLMs are generating and executing code at runtime, the command surface becomes harder to reason about. Traditional assumptions around trust boundaries start to weaken. Peter has been appropriately candid about OpenClaw’s security gaps — and that honesty is healthy for the space. My current view: we need to be careful not to over-optimize for ease at the expense of long-term safety. The winning pattern will likely balance CLI-level appeal with stronger capability controls and auditability. We are still early in figuring out what that equilibrium looks like. Are you comfortable shipping agent systems that lean heavily on CLI surfaces today — or are you seeing the value in MCPs? #AIInfrastructure #AgentSystems #LLMSecurity #HVSays

15

Here are two trends I took away from CES this year: one convergence and one divergence. The convergence is happening around measurement. For nearly two decades, measurement has lived in two separate worlds. On one side, pixel-based, last-touch attribution. Fast, self-attributable, easy to optimize against in near real time, but fundamentally inaccurate and unable to measure true lift or incrementality. On the other, the offline and aggregated world of MMM, incrementality, and lift studies. More accurate, but slower and far less actionable day to day. Advertisers have been forced to juggle both, trying to reconcile conflicting signals and make budget and optimization decisions across systems that were never designed to work together. That gap is finally starting to close. We’re seeing real progress toward convergence, with several measurement vendors developing enhanced versions of CAPI that pass richer attribution and incrementality signals. The promise is better alignment between optimization and true incrementality, and a meaningful reduction in the long-standing disconnect between what’s easy to optimize and what actually drives impact. The divergence is driven by agentic systems. It’s still early, but it’s increasingly clear that agentic is here to stay and will only grow in importance. A key implication is that UI and workflow will become a much less durable competitive advantage across AdTech players, whether DSPs, Exchanges, or Agencies. As agentic systems make it easier and cheaper to build functional, customizable interfaces, more companies will build their own UIs tailored to their specific needs. As a result, the UI layer itself will lose its competitive edge, with differentiation shifting elsewhere. Instead, competition will diverge toward the edges. On one edge, upstream to UI: sales and servicing. Companies that win by acquiring customers efficiently and supporting them exceptionally well. On the other, downstream to UI: infrastructure. Companies that focus on building the best pipes, algorithms, and data-driven orchestration layers. The middle, where workflow and UI once mattered most, is getting thinner.

100

5 Comments

Ever wondered how #AIAgents decipher and know exactly what to do (plus more importantly know what not to do!) when a customer asks a rather vague question. Max Fortis talks through how in the Quiq platform you can leverage Process Guides, Reasoning, Tools (and the most powerful AI for CX tools we know of) to ensure your #AIAgents can handle any request. Just as importantly ensuring that AI Agents know when to pass to their human colleagues on the Quiq platform when the requirement dictates, to combine for the perfect blend of Human and AI expertise.

1

Today’s launch of the Agentic Roadmap today by IAB Tech Lab is more than a vision, it’s a call to action! As agentic execution becomes increasingly part of how digital advertising operates today, this will give the industry a shared playbook for scaling intelligent workflows responsibly and at speed

2

In any organization, time and resources are finite—but distractions are not. When teams align around clear priorities and focus on what truly matters, execution accelerates, quality improves, and impact compounds. Strategic focus is not just about doing less—it’s about doing what drives the highest return. The result? A high-yield dividend in productivity, morale, and long-term value. Revenue growth is a nice side effect. https://lnkd.in/d3m82ipd

50

1 Comment

Doing nothing in payment integrity is costing healthcare billions—and it’s only getting worse. Denials, rework, and outdated recovery models aren’t enough anymore. Codoxo ‘s latest blog shows why the future is pre-claim education and how our Point Zero Payment Integrity solution helps stop errors before they happen, saving time and money. 👉 Read more: https://lnkd.in/gjX_WihV #HealthcareInnovation #PaymentIntegrity #HealthTech #ValueBasedCare #AIinHealthcare #DenialsManagement #PayerProvider #Codoxo #AIproducts

5

LiveRamp's new agentic AI tools connect the full power of our data collaboration network, fueling agents with trusted signals, governance, and interoperability built in. This isn’t just an update, it’s a reset. Now marketers can: ✅ Plan smarter campaigns ✅ Optimize faster ✅ Prove impact everywhere it matters With agentic orchestration, we’re building an open, outcomes-driven ecosystem. Get all the details on this update: https://lnkd.in/gxp2isSG

79

There is an inherent trade-off between reliability vs. flexibility when you build with large language models. Thiaga is one of the architects of Sierra’s Agent SDK, which is the #1 reason agents built on Sierra perform so well and behave so naturally. 1. Agent SDK pushes out the efficient frontier of this reliability vs. flexibility trade-off, allowing you to think about delighting customers while completing complex procedures, as opposed to having to pick one or the other. 2. Agent SDK allows your agent to intentionally and contextually choose its position on that frontier, so it is always acting appropriately. 3. All of this is true whether you’re building in code or no-code, which is all backed by the Agent SDK. Add these up, and it’s why so many CX and Tech teams alike are choosing to build on Sierra right now.

11

The IAB Tech Lab has released a major update to the Open Measurement SDK (OMSDK) that introduces device attestation, a transformative step forward in protecting digital advertising integrity, especially across CTV environments. At its core, this new capability helps the ecosystem verify that an ad was served on a legitimate, unmodified device, rather than an emulator or spoofed environment often used in fraud schemes. By doing so, it closes one of the most exploited gaps in measurement and verification. What makes this update so powerful is its integration of Privacy Pass, an emerging open standard developed through the IETF to enable trust without sacrificing user privacy. Privacy Pass allows a device to prove its authenticity cryptographically, that it is a genuine, attested device, without revealing personally identifiable information or creating a traceable identifier. Here’s how it works: 1. A trusted attester, typically a device manufacturer or operating system provider, issues a privacy-preserving token confirming the legitimacy of the device. 2. The OMSDK integrates these attestation tokens into ad measurement workflows, allowing verification vendors and ad servers to confirm that impressions originate from verified devices. 3. Because the tokens are cryptographically blinded, they cannot be used for tracking, maintaining alignment with global privacy expectations while ensuring trust in the underlying hardware environment. This represents a long-overdue fusion of security and privacy-by-design in advertising measurement, a model that makes fraud far harder to execute while maintaining user trust. However, its success hinges on industry-wide adoption. Device manufacturers, CTV platforms, and OEMs must implement the latest release of OMSDK within their environments. Brands and agencies should demand this capability as a prerequisite for CTV and mobile measurement. Without broad participation, the industry risks leaving the door open to continued fraud. If we want to combat fraud meaningfully, this is the path forward. There is ZERO reason not to adopt the latest OMSDK update. To ignore it would be to signal that we are comfortable with the status quo, and with the billions of wasted ad dollars that come with it. This is exemplary work by the OMSDK Working and Commit Groups for their continued technical leadership, and to Amazon, especially Vinod Panicker, for its stalwart support in architecting and developing this essential update. This is open standards at their best: privacy-preserving, security-enhancing, and built to protect both the consumer and the advertiser. Learn more: Press Release: https://lnkd.in/eTGCeADR Implementation Guide: https://lnkd.in/e3yZhviY

72

1 Comment

🚀 Roku + Amazon Ads + AMD + AWS: A Strategic Tetra-Alliance in CTV & Cloud AI Roku & Amazon Ads Partner on CTV Scale unlocked: Combined reach across Prime Video, Roku Channel & others → ~80 million U.S. homes (> 80% of CTV households) . Performance uplift: 40% more unique viewers, 30% fewer ad repeats, and 3× ROI in early tests . Stock reaction: Roku stock jumped ~9–11% pre-market — validation from investors and analysts . 🔑 Strategic Takeaways: CTV dominance: Roku + Amazon DSP gain premium reach across top-tier streaming platforms — a far cry from traditional TV fragmentation. Revenue diversification: Roku’s platform ad revenue (~$1B+) strengthens gross profit margins. Ad tech shakeup: Competitors like Google and The Trade Desk need to respond with scale and precision offerings. AMD & Amazon: AI Chips & Cloud Collaboration Advancing AI Infrastructure: At AMD's "Advancing AI" event, Piper Sandler raised its price target to $140 following Intel's showcase of the "Helios" server platform and AMD Instinct MI chips — with Amazon Web Services sponsoring the event and hinting at deeper collaboration. AWS Potential Adoption: Reports suggest AWS is evaluating AMD’s MI400 line for future use — aiming to reduce dependence on Nvidia and add computing flexibility. Amazon Stake in AMD: Amazon now holds ~$84M worth of AMD shares — a vestige from AMD’s ZT Systems acquisition . 🔑 Strategic Takeaways: Compute choice: AWS increasingly supports multi-vendor infrastructure — AMD EPYC & Instinct chips gain broader platform adoption alongside AWS’s own Graviton and Nvidia lines . AI scale shift: AMD’s MI400 servers (Helios) aim to challenge Nvidia’s dominance — optimized for open networking and inference workflows. Investor signals: Strong analyst upgrades on AMD highlight potential re-rating as a Tier 1 leaderboard in the AI chip race. 🔎 What to Track Next: Rollout timings: Roku–Amazon integration Q4; AWS/AMD compute services debut window. Ad fatigue measures: Will users accept 30% fewer ad repeats, or will frequency fatigue sink in? Cost/performance comparisons: How AWS’s MI-powered cloud stacks up against Nvidia-based offerings. Stock moves: Could Roku and AMD continue roaring, while ad-tech (TTD) and Nvidia shift? 💬 Let’s Discuss: Are advertisers ready to fully pivot TV budgets to CTV? Will AMD + AWS crack open meaningful Nvidia share? How do these partnerships reshape your investment or business outlook? 👇 Share your thoughts! #Streaming #CTV #AdTech #Roku #AmazonAds #AMD #AWS #AIChips #CloudComputing #TechInvesting #AIInfrastructure #MediaStrategy #Innovation #BigTech #InvestorInsights #Stacquity

2

1 Comment

Public Comment for CoMP started :) Have a look at give feedback! One of the most important initiatives concerning AI impact on publishers and there traffic. We need the chance to control what happens with the content.

7

Had such a great chat with Vivien Lin! She brought so much wisdom and vivid insights to the conversation: 1. We're trying to work with our users, when they make money we make money. We try not to stand on the opposite. 2. From the start we decided to exit China market. We never touched US users. It's all because we are quite self-disciplined and self-regulated. 3. We consider whether this product or whether a feature can truly solve the problem for the user. 4. From 2018, 2019 is always put our top priority to think about how we bring more new user to the crypto industry. It's not like competing the existing user with other exchange. 5. Ideally, the final product should be you don't need to worry about what functions sit in what page. You just go to the AI or the AI assistant to tell him what you want. 6. The difference in the roadmap between crypto and traditional finance is rooted in how the industry got originated. For traditional finance, banking and payment that's naturally where finance comes from than lending than trading. At crypto, the very first need where the industry originated is mining. 7. I think, we are still in a very precious time window especially for exchange it to become a sort of mega platform in near future. 8. My suggestion is always be open minded - open minded to the technology, business model and even all those new ideas. 9. My suggestion to everyone, even if you are a legal, you are a product manager or you are in the operation team, is to study programming, learn how to code. 10. The slides always tell the good stories but you need to get down to the details to understand what's happening. 11. Before all those regulations put into place there has to be someone familiar with how possible the regulation is going to be implemented in the future and we can get prepared beforehand. 12. Without one major hack exchange cannot be a mature exchange. 13. If I know what they are doing I can read their code, I can get a very direct sense about the quality of the project instead of really looking at the powerpoint. Watch the full session here: https://lnkd.in/dGhbXdwr BingX

10

Ever had a CEO or leader ask, “Can we put AI on the roadmap?” Not because it solves a real problem, but because everyone else is doing it. That is where many product leaders are forced to make hard trade-offs between real customer value, rising AI costs, and finite roadmap capacity. This is why the 2026 AI Product Strategy Guide from Mind the Product resonated so strongly. It moves the conversation from hype to discipline. Source: https://lnkd.in/gznKtcph As AI shifts from experimentation to ROI-driven execution, product leaders need rigor in three areas. 1. Budgeting with Economics in Mind • AI costs are ongoing and usage-driven, not just build costs. • Unit economics must include input tokens, outputs, and retrieval. • A practical rule: expected value should be at least three times direct compute cost before making a feature core. 2. A Practical Planning Toolkit Before something earns a roadmap slot, pressure test it: • Start with the problem, not the model. • Build for reversibility and flexibility so you can change models or providers. • Define kill criteria before you build. • Measure success in reduced manual work and real outcomes, not activity metrics. 3. Organizational Design for AI AI systems need explicit ownership: • AI Product Operator to track unit economics. • Context Engineer to manage retrieval quality and data hygiene. • AI QA and evaluation ownership to manage non-deterministic outputs. This is how cost, quality, and value get managed as a system. One simple exercise for your next roadmap review: take your top three proposed AI features: Ask: if a heavy user runs this feature 100 times a day, does our margin on that user go up or down? If the answer is down and you do not have a convincing plan to change that, treat the idea as unfinished. Send it back for redesign or into the experiment bucket. Main takeaway AI is not a feature trend. It is a new cost structure and operating model. Teams that win in 2026 will treat AI as a measurable investment, plan with economic discipline, and design their orgs to run AI products, not experiments.

21

🚨 (Adweek) LLM disruption also played a role in adtech's Q2 downward trend, "...the complexities of navigating the AI boom—which is causing upheaval in how people search and stifling publisher traffic—and adtech companies are finding themselves in a new kind of pressure cooker." 💡 I think a lot about the turbulence B2B tech marketing pros are navigating: ➡️ If traffic from Search to brand.com is collapsing, pipeline and revenue metrics are at risk due to far fewer high-quality lead volumes ➡️ If publisher traffic from Search is collapsing, content consumption signal is diminished making Intent data less reliable for TALs and ABM tactics. 🦺 Immediate solve: direct prospect engagement through high-quality demand gen - back to basics with tried and true, predictable pipeline will steady the ship during this uncertain time. #b2bmarketing #growthmarketing #revenuemarketing https://lnkd.in/dHQAA8bh

18

🚀 Reimagining Product Discovery at Scale As part of our 2025 reset, we ran dual-location discovery workshops for our stakeholders situated across 🇮🇩🇬🇧🇸🇬🇲🇾🇮🇳 to align cross-functional teams around a shared vision. Our prioritization lens🔎 • Impact/Urgency matrix ⚓️ • Strategic themes (time-boxed bets) ⚡️ • OKR alignment 📊 • Client urgency & revenue impact 💰 • Voice of customer across Sales, Ops, CS, and Partners 🏬 What this delivered: 💪🏻 ✅ Data-driven prioritization with an impact-vs-cost view ✅ Stronger alignment across hubs and functions ✅ A clear 2025 roadmap shaped with Engineering from day one ✅ Tighter collaboration rhythms for Product × Tech × GTM If you’re scaling discovery across locations and want to compare notes, happy to share playbook*. Comment - "Discovery", I will be happy to share it #productleadership #productdiscovery #roadmapping #productstrategy #crossfunctional #b2b #saas #orgdesign #uxresearch #engineering #innovation (*Views are my own. High-level methods only; no confidential details.)

19

4 Comments

Most performance issues start with signal loss. Think of Conversion APIs as your campaign’s translation layer turning messy, fragmented signals into clear, actionable insights. CAPIs give you the clarity to prove what’s working in real time. Learn how you can unlock up to 7× more conversions: https://bit.ly/4aDSNAy

13

I've been a believer in streaming data for a long time...but so much of the AI-data conversation has been about static data. That's now changing and real-time data streaming that bakes in Apache Kafka and Apache Flink from Confluent is bringing the context. "The part that we're unlocking for businesses is the ability to essentially serve that structural context needed to deliver the freshest version," Sean Falconer told VentureBeat.

20

Chatted with a friend at RTB House about IntentGPT — very interesting. The idea: AI can derive user intent more precisely from contextual signals alone, no tracking needed. Considering RTB House's innovations in the space, it's definitely worth striking up a conversation about. https://lnkd.in/eY93uiHh

2

The digital advertising ecosystem has long been defined by its complexity. It’s a sprawling Rube Goldberg machine of acronyms—DSPs, SSPs, DMPs—all stitched together by tenuous connections and a massive amount of human brute force. For years, the industry’s answer to this complexity was to build more tools, creating a fragmented landscape where data gets trapped in silos and efficiency decreases. A seismic shift is underway. We are moving from an era of passive tools to an era of [semi-]autonomous action e.g., with Gemini 3.0 which enables action through Agentic AI. This is a fundamental reimagining of how media is bought, sold, and managed.

15

1 Comment