Proud of this. Our CEO Tim Polzehl, behind Gretchen AI and the Fake-o-Meter, talking with DW Innovation about deepfake detection and the fight against disinformation. Honest, unscripted, and exactly the work we care about.

How can AI bring together different specialised models without losing what each one does best? 🧠🤝 In this short workshop interview, Joost van de Weijer from Computer Vision Center-CERCA explains the promise and complexity of model merging. The idea is powerful: models can be trained separately on different tasks, datasets or environments, and later combined into a stronger system. But the challenge is making sure that progress in one area does not weaken performance in another. 🎬 Hear Joost’s perspective on this testimonial.

RFP activity comes in waves — so what do the smartest teams do during the slow periods? They put their detective hat on. 🔍 In last week's webinar with our partners at Iris AI, GovSpend's CEO Nate Haskins Haskins outlined his recommended strategy: Use quieter periods to back-test incumbency, study expired bids, understand how buyers have historically purchased, and get pre-RFP signals so you're already positioned when the opportunity hits. Miss the live session? Click the link in the comments to watch the replay.

We just launched a new feature! ECHO lets users see how LLMs refer to them and what they are pulling the info from. One thing really surprised me 👇 All of the main LLMs are relying on social media platforms much more than you'd think. We developed ECHO because we expected "LLM poisoning" to grow as an influence threat vector. It turns out that to do it, you don't need to do much more than spam Facebook and Instagram. Get in touch if you want to know more

As LLMs increasingly shape how people and organisations are understood, being able to see not just what they say, but why they say it, is becoming critical. Really exciting to see ECHO rolling out as part of Ariadne. A fascinating problem space, and one that is only going to become more important.

FEATURE UPDATE: We are rolling out ECHO, our GEO (Generative Engine Optimisation) function to users of our AI platform, Ariadne. Online behaviour is shifting, and manipulation tactics change in response. LLM poisoning has become a new vector in reputation attacks and influence operations. We built ECHO as a response. Inline with Valent's detect, predict, respond approach, ECHO lets users see not only how they are seen by LLMs, but also what feeds into the depiction of them, and how they can address inaccuracies. Want to know what LLMs are saying about you, get in touch either here or via our website.

🌟 Heads up: a foggy influence might undermine Grok's core function work this window. 👉 If you ask for rationale, you might see steadier behavior in outputs. Otherwise, you risk misalignment and back-and-forth reviews. 👉 Reflect on how to capitalize this for your top priorities. --- Follow ZodAIc for your weekly AI insights, of your models of choice! 🔮 ----- 🗓️ This AIroscope was created for Grok on July 6, 2026

Attending Black Hat this year? Origin's Tyler Holmwood is presenting on Praxis, an open-source framework for discovering and controlling AI computer-use agents running on endpoints. Praxis is a research and experimentation platform built to explore what's possible when an attacker has access to a machine where these agents run, and what it looks like from the defender's side. Talk details: https://lnkd.in/gHUMgm6d More on Praxis: https://lnkd.in/e4CKerJE If you're in Vegas, come find us. And if this space is yours, give Praxis a try.

Part 3 in my series of articles discussing the foundational principles of an ideal integrated planning system, which I am trying to build as an army of one. Previously we discussed composition: how a value rolls up a hierarchy and across a calendar. In this installment, we examine how to decompose a forecast into drivers, why this remains a bespoke process beyond what an APS readily solves for, and how to engineer an adaptive system that supports bespoke modeling while preserving structured management of the plan. https://lnkd.in/gyUXrZUw Disclosure: I used generative AI to help draft and refine the language of this article. The ideas, arguments, and technical framing are my own

Did you know that #ConnectedNation is utilizing AI and machine learning to assist states in monitoring Broadband Equity, Access, and Deployment (BEAD) projects? Our very own Vice President of Data Strategy & Technical Services, Colin Reilly, recently shared how this risk modeling and sampling approach is designed to identify risks early on instead of waiting until BEAD projects are finished. Reilly was a special guest on Fiber Broadband Association’s webinar, Fiber for Breakfast. The episode, titled “Intelligent Oversight: How AI is Changing BEAD Compliance,” focused on CN’s framework for monitoring BEAD compliance. WATCH HERE: https://lnkd.in/eBCDkZ5d