Monday was Labor Day. On Tuesday, enterprise software went right back to manufacturing labor. Somewhere between Identity Week in DC and a train ride from Union Station to Princeton, I found a much simpler way to describe the last-mile identity automation we do at @redblockAI: We make identity tickets unnecessary. ServiceNow says more than 100 billion workflows run on its platform every year. How many of them end by asking a human to finish the work? A lot. A ticket lands in a queue. Someone picks up the phone, sends a few emails, opens two or three admin consoles, completes the task, takes screenshots, uploads them to the ticket, and marks it complete. The screenshot is important. It proves to the software that the human followed its instructions. The ticket was created by software. Everything was automated. Except the work. A ticket is often just software saying: “Here’s what needs to happen. Someone else please do it.” That is a strange definition of automation. Keep the ticket if you need a record. But let’s stop making it someone’s job. Today, we’re planting a flag at Redblock: NO TICKETS. Bring us your most hated identity ticket. We’ll show you how to make it unnecessary.

We started building on open weights before it was cool and ran a POC in a customer's cloud in 2024. This week @redblockAI co-signs the Open Weights Letter. Alongside NVIDIA, Microsoft, Meta, Google, OpenAI, IBM, Hugging Face, the Linux Foundation, Harness, and others. For us it is not about AI ideology. It's about how enterprise AI should actually be deployed. We build AI that executes inside some of the most sensitive identity infrastructure in the world. When AI is running privileged production work, customer data shouldn't leave their environment for live inference. That's why we deploy open weight models directly inside our customers' private tenants and on-premises environments. The inference happens where the data already lives. Customers already run their identity systems, databases, and Kubernetes clusters inside their environment. AI performing critical work shouldn't be the exception. This hits close to home for me. I started my career building on Linux, Apache, PHP, and MySQL. Open source shaped how I think about software architecture. Over three years ago, I wrote that enterprise AI would eventually need to run inside customer-controlled environments. And today, that shift is officially reality.

yes, @redblockAI has redblock for redblock. Identity agents for a SCIM-less world.

Big day for us at @redblockAI. We’re part of @SailPoint’s universal platform access launch.

one of the biggest unlocks for AI is how it shrinks time. in this instance we built a feature that shortens rollout of identity governance products. what used to take 4 years, can now be done in 2. massive parallelization of efforts using agents.

Today we launched yet another feature to beat the crap out of manual work in identity security.

Good fun recording a new episode of @IDACPodcast with Jim and Jeff today at #GartnerIAM in Dallas.

Super pumped to share this publicly. 📢 Redblock's Agentic AI brings data from disconnected apps onto @SailPoint. What used to take days--CSV files, manual data-wrangling--gets done in 2 minutes! @redblockAI

⚡️New pixels for @redblockAI in the Identity Governance market map ⚡️ Thanks @InvestiAnalyst 🙏

Winners from yesterday’s #TakeTheAIChallenge at #RSAC

Sneak peek of #TakeTheAIChallenge at #RSAC

Ready to be wheeled for #RSAC. Air-gapped Identity Security flows. neon lights=yes fan speed=brrrr; full yank_weights=yes memory=yes compiled_ui=yes pre-authenticated=yes cheap_nvidia=yeah model_count=2 tired=no Beast mode in our office.

Treating AI agents as ordinary “non‑human identities” (NHI) guarantees a security disaster. Agents can reason, re‑plan, pivot across systems, and ask for fresh privileges throughout one workflow. Almost human-like, or "super-human" capabilities.

🚨 Big news: @redblockAI now integrates with @saviynt. Our Agentic AI connector brings automated user lifecycle management to hundreds of disconnected apps—right inside the Saviynt Identity Cloud. Here’s why it matters:

🚨 Take the AI Challenge at RSAC 2025! Can you guess how fast AI completes Identity tasks? Win one of 25 "Turing Machine" board games—perfect for cryptography enthusiasts! Here's why👇 #AI #RSAC #CyberSecurity @RSAConference

Our new open source drop on @github. MCP -- meme creation platform. 😂 Give it a spin. Clone the repo. Get your memes from @imgflip using AI and @AnthropicAI Demo vide below. Link in the first comment. @redblockAI

📣 We’re happy to introduce our latest benchmark, the 𝗣𝗔𝗥𝗥𝗢𝗧-𝟯𝟲𝟬𝗩! 🚀 PARROT-360V is designed to see how well 𝗩𝗶𝘀𝗶𝗼𝗻 𝗟𝗮𝗻𝗴𝘂𝗮𝗴𝗲 𝗠𝗼𝗱𝗲𝗹𝘀 (𝗩𝗟𝗠𝘀) can handle complex, multi-step puzzles, requiring them to process both text and visual clues—similar to real-world challenges. What started as a simple internal experiment to benchmark language models before we push to CI/CD became a side-project. 😋 A month ago, we released dual datasets for text, and now 2300 samples of Jumble puzzles for Vision tasks. I’m totally amazed. All of us are getting started! Frontier models and us developers helping models push the limits. Next stop? A cybersecurity dataset. Kinda easy guess—where this is all going 😁🙏 #ai @redblockAI