Answers from the documentation; checking which model is on duty
Hi - I answer from the OpenSmartRoute documentation: routing, the API, plans and quotas, self-hosting. Ask away, or open a support ticket if you need a person.
Grounded in the docs - follow a source before acting on it.
TechCrunch Disrupt 2026 Day 2 Covers Robotics, Security, and Enterpris - OpenSmartRoute
TechCrunch Disrupt 2026 Day 2 Covers Robotics, Security, and Enterprise AI
Day 2 of TechCrunch Disrupt 2026 features sessions on robotics, agent security, and the shift from chatbots to physical AI systems. Speakers discuss funding, hardware, and the new rules for building AI in the enterprise.
Key points
A $100,000 equity-free funding round is available for the Startup Battlefield.
TechCrunch Disrupt 2026 connects 10,000+ founders, investors, and technology leaders.
Sessions cover humanoid robots, storage maximization, and agent security.
Benchmark announced a full partnership with the event organizers.
Why it matters: Engineers and managers learn about new robot capabilities, security risks for agents, and how to fund their AI projects.
By OpenSmartRoute editorial · written through the router by writer-small
From TechCrunch AI - “TechCrunch Disrupt 2026: Day 2”
Day 2 of TechCrunch Disrupt 2026 took place at Moscone West in San Francisco. Teams gathered to explore startups and solution providers in the Expo Hall. They used AI tools to connect with over 10,000 founders and investors. The event focused heavily on artificial intelligence and its real-world applications. Speakers from major companies shared their latest ideas on robotics and security.
The day featured sessions on how physical AI systems are changing industries. People discussed funding strategies for new hardware and software ventures. Leaders explained the new rules for building safe and effective AI in enterprises. Attendees could see products in action and speak directly with the teams behind them.
Day 2 Overview - What happened at TechCrunch Disrupt 2026 on the second day
The second day of the conference ran from 8 a.m. to 5 p.m. in the Level 1 lobby and Level 2 Expo Hall. Visitors picked up badges and registered in person to access the venue. The schedule included live startup pitches and fast-paced presentation rounds.
Speakers covered topics ranging from robotics to public safety and venture investment. The agenda listed specific talks on AI hardware and the future of physical products. Teams could explore 300+ startups alongside tech partners and solution providers.
Registration hours were strict, running only until 5 p.m. Pacific Time. Badge pick-up required being present at the Level 1 lobby. All times listed in the schedule followed Pacific Time standards.
Attendees could choose between an Expo+ Pass or an eligible conference pass. The Expo+ Pass focused on discovery and networking within the Expo Hall. Other passes added stage sessions, interactive roundtables, and Q&A breakouts.
The event aimed to help people find their next investment or business solution. Leaders from Nvidia, Waabi, and Flock Safety gave keynotes. Benchmark announced a full partnership with the event organizers.
People could use AI-powered matchmaking to connect with a community of founders. This tool helped schedule meetings and follow up on introductions. The TechCrunch Events app facilitated these connections during the conference.
Key Themes - The main topics like robotics, public safety, and venture investment
Dell and Intel report that enterprise AI budgets are burning through cloud tokens too fast. Companies are moving toward on-premises infrastructure and open-weight models to save money.
Mirror Particle raises capital to create an AI engine that simulates changing human motivations. The company rejects large language models in favor of a foundation model trained on longitudinal data.
Robotics and hardware took center stage on the second day of the event. Sessions discussed humanoid robots and edge AI technologies. Speakers showed how physical products are moving from prototypes to production.
Public safety was another major theme, especially regarding AI surveillance. Flock Safety presented ideas on using AI for security and safety monitoring. The future of public safety involved integrating AI into existing infrastructure.
Venture investment discussions focused on funding models and startup growth strategies. The Startup Battlefield competition offered $100,000 in equity-free funding. Investors compared perspectives on what makes a business fundable.
Security and enterprise topics highlighted the complexity of the cloud environment. Speakers addressed the agent security problem and SaaS business models. Companies debated how AI breaks traditional software business models.
The shift from chatbots to physical AI systems was a recurring discussion point. Experts asked what comes after the chatbot era. The focus moved toward agentic AI and decision-making capabilities.
Open-weight models and the new AI stack received significant attention. Mercedes Bent and Vanessa Larco from Premise spoke on this topic. Teams explored how open models fit into modern infrastructure stacks.
Startup Battlefield - Details on the $100,000 equity-free funding competition
The Startup Battlefield competition continued with healthtech and life sciences categories. Enterprise and security sectors also competed for a place in the final round. The prize pool included $100,000 in equity-free funding.
Founders could sharpen their plans based on investor feedback. Investors compared perspectives on what makes a business fundable. Startup teams understood the priorities shaping their next stage of growth.
The competition featured talks on Series A funding for 2027. Jahanvi Sardana, Index Ventures, and others discussed funding trends. Yuri Sagalov from General Catalyst shared insights on non-AI startups.
Build Mode Live covered building from zero to one. Juliana Di Simone from Forela and others presented their growth strategies. The session helped teams understand the challenges of early-stage growth.
Revealing AI Disruptors 60 explored what AI can do next. Parag Agrawal from Parallel Web Systems and Shay Grinfeld from Greenfield Partners spoke. They discussed the capabilities and limitations of current AI systems.
The Real Tokenmaxxing session examined how companies navigate a multi-model world. Mo Jomaa from Capital G and Vipul Ved Prakash from Together AI presented. They analyzed the economic strategies behind AI adoption.
Building the Next Generation of AI Giants featured Jas Khaira from Blackstone N1. The talk covered the scaling challenges of AI infrastructure. The Zero-to-1K Playbook offered strategies for acquiring first customers.
Crystal Huang from GV shared how to get 1,000 customers without a marketing budget. Grant Lee from Gamma and Elia Wallen from Engine provided practical advice. These strategies helped startups reduce customer acquisition costs.
So You've Got a Hit Product addressed how companies repeat their success. Maddi Holman from Daring Ventures and Filip Kaliszan from Verkada spoke. They discussed product scaling and operational efficiency.
Tamar Yehoshua from Atlassian joined the conversation on product leadership. The discussion covered how to maintain momentum after a successful launch. Attendees could take fresh ideas back to their own teams.
Robotics and Hardware - Sessions on humanoid robots, edge AI, and physical products
Building the next generation of AI giants involved creating physical products. Sessions explored practical applications and emerging markets for robotics. Challenges of deployment in real-world environments were a key focus.
A live robot demonstration showed robots built for real life, not demo reels. Aaron Edsinger from Hello Robot presented Stretch, a robot designed for utility. The robot could perform tasks outside controlled environments.
Robots were waiting for their ChatGPT moment, according to Les Karpas from NVIDIA. He explained what was standing in the way of widespread adoption. Hardware limitations and cost remained significant barriers.
Operating at the edge meant AI worked when the cloud failed. Ali Agha from FieldAI and Michelle Lee from Medra discussed edge computing. This approach reduced latency and improved reliability for critical systems.
Launching AI's orbital infrastructure involved space applications. Devin Coldewey and Blake Bullock from The Aerospace Corporation spoke. Jeff Thornburg from Portal Space Systems and Jonny Dyer from Muon Space Systems joined the discussion.
Humanoid robots arrived live and on stage. Jonathan Hurst from Agility Robotics and Digit, a humanoid robot, were featured. The session showed how prototypes are moving toward production.
From prototype to production raised questions about scaling in reality. Kevin Peterson from Bedrock Robotics and Adrian Macneil from Foxglove addressed this. They discussed the engineering hurdles of mass production.
The race to general-purpose robots involved Skild AI. Varun Jain from SE Ventures and Deepak Pathak from Skild AI presented. They explained how AI meets the physical world in new ways.
Can we engineer nature's comeback? Ben Lamm from Colossal Biosciences asked this. The session explored biological applications of AI technology. This topic bridged robotics with biotechnology and environmental science.
Why and how every company will build its own AI was another topic. Manos Koukoumidis from Oumi and Matt Burns from Insight Media Group spoke. They discussed the shift toward internal AI development capabilities.
Security and Enterprise - Discussions on agent security, SaaS models, and cloud complexity
The agent security problem nobody is talking about received serious attention. Gavriel Cohen from NanoCo and Ric Smith from Okta presented their findings. They highlighted risks in autonomous agent systems.
Rewriting SaaS challenged the old business model. Cathy Gao from Sapphire Ventures, Aaron Jacobson from NEA, and Barr Moses from Monte Carlo discussed. They explained how AI changes software delivery and pricing.
Before letting AI do a job, teams must verify its ability. Irene Solaiman from Hugging Face emphasized the need for rigorous testing. Evaluation processes needed to match the complexity of AI tasks.
The $63 billion case for storagemaxxing your AI stack involved cloud infrastructure. Pinaki Mukherjee from Wasabi Technologies, Stephanie Walter from HyperFRAME Research, and Kevin Cochrane from Vultr spoke. Jake Lee from L2 Point Management joined the discussion.
What comes after the chatbot era remained a pressing question. Alexander Embiricos from OpenAI addressed the future of conversational AI. The focus shifted toward agentic workflows and autonomous decision-making.
The GTM engineer role emerged as a new job category. Kareem Amin from Clay explained how AI created this position. Tech companies needed engineers who understood both growth and AI.
Model behavior sessions showed how top startups get real value from adoption. Cat de Jong from Anthropic, Grant Lee from Gamma, and Kareem Amin from Clay presented. They analyzed metrics and outcomes of AI implementation.
Proxy, host, operator discussed who can read an agent's prompts. Richard Rizk from BLACKBOX AI highlighted the importance of prompt visibility. Security teams needed to understand the full data flow.
Securing the AI enterprise made the cloud more complicated. Katie Moussouris from Luta Security, Wendy Nather, and Rudy Mitra from AWS spoke. They addressed the growing complexity of enterprise security landscapes.
The enterprise isn't broken, but assumptions about it are. Arsalan Tavakoli from Databricks challenged common beliefs about enterprise systems. Teams needed to update their mental models of AI integration.
Use the audience Q&A to explore decisions you are facing. Attendees could ask how ideas applied to their specific businesses. Better understanding implications helped companies evaluate potential partners.
Why it matters - How this news impacts cost, speed, quality, and safety for model runners
This news matters because it signals a shift from chatbots to physical AI systems. Engineers must prepare for hardware integration and edge computing needs. Managers need to understand the new cost structures of agentic AI.
Security risks are increasing as agents gain autonomy. Teams must implement guardrails before deploying autonomous systems. Safety becomes a critical factor in enterprise adoption decisions.
Speed of deployment changes with edge AI and local models. Latency issues disappear when processing happens closer to the data source. Cost savings emerge from avoiding expensive cloud calls for simple tasks.
Quality of outcomes improves when models are evaluated on real prompts. Evaluation tools help teams measure routing accuracy and model performance. Teams can fail builds if evaluation scores drop below thresholds.
Safety concerns extend beyond data privacy to physical world interactions. Robotics and autonomous agents require rigorous testing protocols. Regulations may emerge to govern physical AI deployments.
How OpenSmartRoute helps - Which router features apply to the security and agent topics
Teams that route requests through OpenSmartRoute gain better control over agent security and costs. The router scores every candidate on quality, cost, speed, and safety. The team sets the weights per request to prioritize specific factors.
Hard rules pin requests to specific models based on data sensitivity. Text with personal data stays on an on-premises model. A region or a cost cap is never crossed by the routing logic.
The input guard spots prompt injection and personal data before a request leaves. This feature protects sensitive information from unauthorized access. It prevents data leaks in enterprise environments.
The hosted platform keeps a models catalogue with prices and public rankings. These rankings are built from real traffic data. Teams can see which models perform best in practice.
An input guard spots prompt injection and personal data before a request leaves. This feature protects sensitive information from unauthorized access. It prevents data leaks in enterprise environments.
The savings ledger shows what each routed request cost next to the most expensive model. Teams can track cost reductions over time. This data helps justify AI investments to stakeholders.
'osr eval' measures routing accuracy on the team's own prompts. The tool can fail a build when it drops below a threshold. This ensures the routing logic remains reliable and accurate.
A new model is one catalogue entry and competes on the next request. Nothing else changes in the app when a new model is added. This makes experimentation fast and low-risk for teams.
OpenSmartRoute works with any OpenAI-compatible provider and open-weight models. It supports local serving for sensitive data processing. The router integrates with MCP tools and A2A agents seamlessly.
What to do - Practical steps for attendees to plan their visit or research
Attendees should register for a pass before the event starts. The walk-up rate is 50% off the original price. Get your pass and join the conference to avoid missing sessions.
Check the full agenda for session times and the complete Day 2 lineup. Note the companies and approaches that stand out during the fast pitches. Take fresh ideas back to your team after the event.
Meet the founders afterward to explore potential opportunities. Use the networking lounge to connect with relevant people. Follow up on an introduction and agree on a next step.
Compare investment perspectives with other attendees afterward. Consider what the discussion means for your business or portfolio. Bring a business challenge to small-group discussions.
Explore the ideas, business strategies, and technologies shaping the startup ecosystem. Use the TechCrunch Events app to schedule meetings. Connect with people relevant to your goals.
Research the $100,000 equity-free funding competition details. Understand the priorities shaping the next stage of growth. Compare perspectives on what makes a business fundable.
Visit the Expo Hall to see products in action. Discover tools that could help your business grow. Speak directly with the teams behind the products.
