LinkedIn image ad examples

It's always energizing when one of our founders helps move an industry conversation forward. This week, TechCrunch's Russell Brandom dug into one of the most-hyped (and least-understood) ideas in AI right now: recursive self-improvement (RSI), the point at which AI systems can meaningfully improve themselves. For the piece, he turned to Doris Xin, Co-Founder & CEO of .406 portfolio company Disarray. Doris doesn't see #RSI the way most of the field does. The common view treats it as a problem of intelligence, that once models get smart enough, they'll start improving themselves. Her argument is that the intelligence is largely here already, and what's missing is the infrastructure to put it to work: "Given infinite compute and infinite time horizon, we are already there. I want to make an argument that this is not a creative endeavor, really. It's just a lot of meat-and-potatoes engineering." It's a characteristically clear-eyed take from someone who builds rather than speculates. She demonstrated the value of infrastructure to activate the current model intelligence in her recent 24-hour run, where on a single A100, Disarray's MLE agent earned 28 Kaggle medals, with nine top-10 finishes and a #1 ranking ahead of every human team. Congratulations to Doris and the team on the well-deserved recognition. You can read the full article below. https://lnkd.in/eEt3fVk2 Moustafa AbdelBaky, Kevin C. Wang

In a TechCrunch article on RSI, I was quoted saying “[RSI] is not a creative endeavor.” What makes me say this? To oversimplify, RSI is about finding configurations that lead to AI improvement from a vast, uncertain space. With infinite resources, you can sample randomly until something works. Without infinite resources, progress depends on testing high-quality hypotheses. That is where “creativity” comes in. A creative breakthrough often means sampling from a subspace no one had thought to search before. AI can test far more hypotheses than humans, but it can’t generate good, novel hypotheses effectively if it doesn’t understand the entire search space and can’t leverage existing knowledge to prune the space for efficiency. The "meat and potatoes engineering" I was referring to is about building specific mechanisms to overcome these limitations: A context graph tells an agent what has already been tried, which evidence is trustworthy, which assumptions are stale, and where adjacent opportunities may exist. A harness keeps the search process coherent over time, so the agent does not lose state, repeat failed work, overtrust its own judgment, or keep pursuing a bad trajectory. These are the hard engineering challenges we need to solve to meaningfully move towards RSI today. The full TechCrunch article by Russell Brandom offers diverse perspectives on the topic from luminaries in the field and a sharp synthesis: https://lnkd.in/gWy6qqas

DKube turns 500-page loan files into structured data in minutes.

DKube reconciles mortgage documents in minutes at 99.6% accuracy — fully in your private environment.

AI infrastructure isn’t just hard — it’s the #1 drag on performance and ROI. The 2026 State of AI Infrastructure Report reveals how winning teams simplify, unify, and accelerate their AI foundation.

𝗪𝗵𝗮𝘁 𝘀𝗲𝗽𝗮𝗿𝗮𝘁𝗲𝘀 𝗔𝗜 𝗹𝗲𝗮𝗱𝗲𝗿𝘀 𝗳𝗿𝗼𝗺 𝗲𝘃𝗲𝗿𝘆𝗼𝗻𝗲 𝗲𝗹𝘀𝗲? The new 2026 report reveals the patterns. Download today.

Cada día sin Deepdub es un día con llamadas que podrían sonar mejor, costar menos y vender más. Deepdub hace lo que otros prometen: 📞Voces indistinguibles de una persona real. 📞Latencia de 125 ms, imperceptible para el usuario. → Escalabilidad ilimitada sin perder calidad. 📞Cumplimiento SOC 2 y GDPR para el sector financiero europeo. No es casualidad que Google, NVIDIA, AWS y Microsoft hayan elegido a Deepdub como socio estratégico de voz con IA.

Rethinking the Future of AI: It’s All About the Data Hear Ari Morcos, CEO of DatologyAI, on the Latent Space podcast to discuss the importance of data in the current industry. Ari describes how “models are what they eat”—and that careful curation: filtering, rebalancing, sequencing, and generating synthetic data is the key to training smaller models to better performance, all with less compute. His own journey, from focusing on model architectures to realizing the outsized impact of data quality, highlights how curation breaks through the diminishing returns of naive scaling. At DatologyAI, the mission is clear: automate this complex process to make world-class curated data accessible to everyone, ushering in a new era where data efficiency, not brute force scale, drives AI progress. Interested in joining our team and solving the world’s most interesting data problems? Apply here: https://lnkd.in/gz-unQ48 Watch the podcast here: https://lnkd.in/gK6c_dRu

Mehrere führende Sprachmodelle in einem Tool, gehostet in Deutschland, direkt in Ihren Office-Programmen. Holdinggesellschaften wollen KI für alle Abteilungen einführen, ohne die bestehende Microsoft-Umgebung anzufassen. Sie brauchen ein Tool, das alle nutzen, und Daten, die das Land nicht verlassen. Genau dafür ist DeutschlandGPT gebaut. Wer hier zu lange wartet, sammelt weiter Einzellösungen ein, die später teuer zusammengeführt werden müssen. Das passiert nicht irgendwann. Das passiert täglich. In deutschen Unternehmensgruppen jeder Größe. Über 34 Sprachmodelle in einer Plattform, direkt als Add-In in Ihren Office-Programmen, gehostet in Deutschland. Die kostenlose PDF gibt Ihnen die Orientierung dafür: ✔ Welches der über 34 Modelle sich für welche Aufgabe eignet, inklusive Auswahlhilfe ✔ Wie KI direkt im Microsoft Stack integriert wird ✔ Warum BSI C5 und Zero-Data-Retention den Unterschied zu US-Tools machen ✔ Der ehrliche Vergleich zu einer reinen Microsoft-Umgebung Jetzt kostenlose PDF herunterladen.

Wissen Sie, wo die Kalkulation liegt, die Ihr Controlling letzte Woche von einer KI hat gegenprüfen lassen? Wahrscheinlich nicht. Denn wenn Mitarbeitende Dokumente in ein privates KI-Tool kopieren, liegen sie Sekunden später auf Servern, über die niemand im Haus Kontrolle hat. Kein Einzelfall: In vier von zehn Unternehmen läuft KI bereits über private Zugänge, die Nutzung hat sich in einem Jahr verdoppelt. Gesperrt ist sie in den meisten Häusern trotzdem. Denn ein Verbot stoppt nicht die Nutzung. Nur die Sichtbarkeit. Und während offiziell nichts passiert, verlassen vertrauliche Daten das Unternehmen – meist Richtung USA. Beim nächsten Audit wird aus dem blinden Fleck ein Problem. Es gibt einen anderen Weg: alle führenden KI-Modelle in einem Tool, gehostet in Deutschland, DSGVO-konform und ohne dass Ihre Daten das Land verlassen. Wie Sie dorthin kommen, haben wir in eine PDF gepackt. 13 Seiten aus über 100 Einführungsprojekten im Mittelstand. Was in der PDF steht: ✅ Eine KI-Richtlinien-Vorlage, die Sie sofort anpassen können ✅ Plattformvergleich: deutsche Server, EU-Cloud, On-Premise ✅ Ein Champions-First-Rollout in vier Wochen ✅ Ein ROI-Rechner für Ihre Unternehmensgröße 👇 Gratis herunterladen

𝗦𝗽𝗲𝗮𝗸𝗲𝗿 𝗔𝗻𝗻𝗼𝘂𝗻𝗰𝗲𝗺𝗲𝗻𝘁: 𝗠𝗼𝗿𝗶𝘁𝘇 𝗛𝗲𝘂𝗯𝗲𝗿𝗴𝗲𝗿 & 𝗖𝗵𝗿𝗶𝘀𝘁𝗶𝗮𝗻 𝗞𝗻𝗲𝗯𝗲𝗹 The need for digital sovereignty and the conscious integration of technological innovations, especially Artificial Intelligence, is widely recognized today. But while the "why" has been thoroughly analyzed, the "how" remains the decisive open question for the European technology sector. As an AI company that has operated independently in this field for eight years, we want to provide a space for concrete answers. At the upcoming 4th dida conference, we are hosting a breakfast session focused on the topic of 𝗗𝗶𝗴𝗶𝘁𝗮𝗹 𝗜𝗻𝗱𝗲𝗽𝗲𝗻𝗱𝗲𝗻𝗰𝗲, using AI as a primary example. The discussion will be initiated by opening impulses from: * Dr. Moritz Heuberger, Member of the German Bundestag (Committee for Digitalization and State Modernization), who holds a PhD in public administration digitalization and is a former project manager at the Federal Ministry of the Interior (BMI). * Dr. Christian Knebel, founder and former CEO of the e-government company publicplan GmbH, now an independent consultant for open-source solutions in the German administration. We are excited to host this event and very much look forward to a constructive discussion with our guests. 📅 June 19, 2026 | Frizzforum Berlin 📩 This breakfast is invite-only. However, if you are an expert in digitalization and would like to join the conversation, please check in with Bela Baganz. You can reach out via LinkedIn or email at [email protected] Dr. Moritz Heuberger Fotocredit: Nils Leon Brauer

𝗦𝗽𝗲𝗮𝗸𝗲𝗿 𝗔𝗻𝗻𝗼𝘂𝗻𝗰𝗲𝗺𝗲𝗻𝘁: 𝗝𝗮𝗻𝗻𝗶𝘀 𝗖𝗵𝗲𝗺𝘀𝗲𝗱𝗱𝗶𝗻𝗲 (𝗧𝗨 𝗕𝗲𝗿𝗹𝗶𝗻) Next week the dida conference returns to Berlin. Diffusion reshaped image and video generation. Could it do the same for language? That's the thread Jannis Chemseddine from Technische Universität Berlin picks up at this year's dida conference. In his talk "Advances in Continuous Diffusion Language Models," he digs into the continuous formulations pushing diffusion into territory long held by autoregressive LLMs, and what they unlock along the way. 📅 June 19, 2026 · 09:00–18:00 CET 📍 frizzforum, Friedrichstraße 23, Berlin-Kreuzberg Full lineup and registration 👉 https://dida.do/conference Few spots left. #MachineLearning #DiffusionModels #LanguageModels #GenerativeAI #didaconference

"We replaced 70% of our agents. Customers never noticed." That's not a product claim. That's a measured outcome from Notch — an AI agent platform serving global enterprise clients across industries. The fix wasn't a new AI model. It was the voice layer. ~125ms response time. 80 emotion styles. 100+ languages, all native. If your voice automation isn't containing calls — this playbook shows you exactly why, and what the solution looks like in production. 👇 Free. One click. Yours.

𝗛𝗼𝘄 𝘁𝗼 𝗴𝗲𝘁 𝘁𝗵𝗲 𝗰𝗼𝗿𝗿𝗲𝗰𝘁 𝗮𝗻𝘀𝘄𝗲𝗿 𝗳𝗼𝗿 𝟯𝟱𝘅 𝗹𝗲𝘀𝘀 𝗶𝗻𝗳𝗲𝗿𝗲𝗻𝗰𝗲 𝗰𝗼𝗺𝗽𝘂𝘁𝗲. Most of the work on making models cheaper to run is really about making the model smaller, like distillation, quantization, pruning, and speculative decoding. They all work, and they're all chasing the same number: the cost of a single token. But the token cost is only half the bill. The other half is how many tokens the model burns to get to an answer, and almost nobody optimizes that. We tend to treat answer length as just how a model responds, not as something you can change. It turns out you can change the response, and it makes a huge difference. Our new paper, Brevity is the Soul of Inference Efficiency, shows that length comes from the data a model is trained on, so you can curate for it. Feed it pretraining data curated for brevity and correctness, and your model will learn to answer in far fewer tokens. Here's what that looks like on a 4B vision-language model: The model reaches a correct answer with 35× less compute than a same-size, same-task baseline and stays within ~1 pp of that baseline's accuracy. The whole gap is about length. Our model answers in about 30 tokens; the baseline takes around 1,284. Self-hosted, that's roughly $1.34 versus $47 per million correct answers. And no, shorter doesn't mean worse. When you match the output length, the curated model comes out +17.55pp more accurate. You don't pay with quality by having a shorter answer. So it's not just cost per token we should be looking at. The number you were not paying attention to that is blowing up your bill is the cost per correct answer, and you can do something about it.

Your forecast was accurate. By the time it reached a planner, conditions had already changed. That gap between prediction and action is where supply chains lose time, money, and ground. This new ebook walks you through how agentic AI closes that gap across the full supply chain lifecycle. Forecasting is the first domino. Agentic AI is the second.

Most enterprise AI platforms are one-size-fits-all and slow to deploy. Cerebro is different: custom-built, live in days, and designed for how your business operates.

Alisa Cohn has coached founders at Etsy, Venmo, Foursquare, and DraftKings. Ask her Delphi anything about leadership.

Ever wondered how the best CEO's get better? They get coached by Alisa Cohn.

Big news for d-Matrix and the future of AI inference. d-Matrix is collaborating with NVIDIA to bring our next-generation Raptor inference XPUs into NVIDIA’s rack-scale AI infrastructure with NVIDIA NVLink scale-up interconnect. As an NVIDIA NVLink Fusion partner, we’re developing Raptor to integrate directly into the NVIDIA MGX rack architecture, creating a scalable foundation for ultra-low latency, premium-level token services. This brings together d-Matrix’s memory-centric inference architecture with NVIDIA’s widely deployed AI factory ecosystem, giving AI labs, hyperscalers, and neoclouds greater flexibility to deploy the right compute for each phase of inference. From AI coding assistants to real-time chatbots and voice agents, the next generation of AI demands faster, more responsive inference at scale. Raptor is being built for that future. Read the full announcement: https://lnkd.in/gsBcAwqC #NVLinkFusion #AIInfrastructure #AgenticAI #dMatrix

Trendspek accelerated their infrastructure inspections 8× using Datature’s easy-to-use AI platform for crack detection. No complex coding or heavy team investment needed - just fast, accurate insights at scale. Even in 3D. Learn how you can simplify your inspection workflows and protect your infrastructure with Datature.

Datature empowered MacroInsight to rapidly train vision AI models for radiology workflows, accelerating smart healthcare innovation. Clinicians now make faster, more informed decisions with secure, and accurate ML-driven insights.

𝐏𝐨𝐰𝐞𝐫 𝐘𝐨𝐮𝐫 𝐀𝐈 𝐖𝐨𝐫𝐤𝐥𝐨𝐚𝐝𝐬 𝐰𝐢𝐭𝐡 𝐁𝐥𝐚𝐳𝐢𝐧𝐠 𝐒𝐩𝐞𝐞𝐝! Looking for unmatched AI performance at unbeatable pricing? Get access to NVIDIA H100 SXM GPUs on Cyfuture.ai – the ultimate choice for training massive models and running high-speed inferencing. ✔Just $2.34/hr ✔Perfect for LLMs, GenAI, and HPC tasks ✔Scalable, reliable, enterprise-ready infrastructure Whether you're a researcher, developer, or enterprise AI team — Cyfuture.ai brings GPU firepower to your fingertips. 👉 Start building with H100 now – no waitlist, no hassle!

An AI factory’s power lies in fast, continuous iteration. The AI-ready data stack drives nonstop feedback loops ,compounding accuracy, accelerating improvement, and scaling enterprise AI systems. Get a demo to learn how it works.

AI factories run on high-quality data pipelines. Dataloop transforms unstructured inputs into AI-ready datasets, automating ingestion, curation, clustering, and querying to power scalable, production-grade AI systems. Get a demo to learn how it works.