Devdutt @devdutt

A Hero That Trusted His Intuition. The problem wasn’t that Morgan Stanley didn’t believe him. Rick Rescorla had already been right once. After the 1993 basement bombing of the World Trade Center, the firm’s security chief walked the towers, studied the site, and reached a conclusion most people preferred not to hear: the World Trade Center was still a target. Next time, he said, it would not be a truck. It would be an airplane. An airplane. In a 1993 report. Executives listened. Then they looked at the lease. Morgan Stanley was the largest tenant in the complex—twenty-two floors of the South Tower, thousands of brokers, and a trading infrastructure that could not be packed into boxes and wheeled across the Hudson. They literally made it a trading center. The contract with the Port Authority ran until 2006. Walking away from premium Manhattan space of that scale would have meant penalties measured in the hundreds of millions, plus the operational chaos of relocating an entire downtown machine. So the company stayed. Rescorla adjusted. If they could not leave the building, they would learn to escape it. He built a protocol and treated it like doctrine. Unannounced drills. No exemptions. Brokers who billed by the minute were pulled off their desks and sent into the stairwells two by two, starting from the top floors, leaving a lane for firefighters coming the other way. People complained. He did not care. Practice, he believed, was the only thing that would beat panic. On the morning of September 11, 2001, American Airlines Flight 11 hit the North Tower. From the 44th floor of the South Tower, Rescorla could see the fire. Then the Port Authority came over the speakers: the South Tower was secure. Stay at your desks. Rescorla grabbed a bullhorn and ignored them. He started the drill his people already knew by heart. Floor by floor, two by two, down the stairs. He sang as he moved—Cornish songs from childhood, then “God Bless America”—keeping the line moving, refusing to let the stairwell become a stampede. Seventeen minutes later, the second plane struck the South Tower. By then, most of Morgan Stanley was already on the way out. Of the 2,687 employees in the building that morning, all but 13 survived. It was stunning. Rescorla could have been outside with them. He went back up instead. Colleagues told him to leave. He said he would, as soon as he was sure no one was left behind. He was last seen around the 10th floor, heading higher. The South Tower fell at 9:59 a.m. They never found him. He was one of the 13. He was a hero in every sense of the word. Remember Rick Rescorla, he remembered for us.

@sonicaghi Zoom zoom!

I wrote about the "New Dawn" in enterprise software. It looks a lot like the cloud and API sprawl era we saw a decade ago, where adoption comes first and governance is an afterthought. But this time with AI workloads it's more dangerous and expensive. Some numbers from our platform that tell you what's really going on in the enterprise world: ▪️ Enterprise AI traffic as a share of total Kong platform traffic is up 7x in nine months. We're processing trillions of tokens, heading into quadrillions. ▪️ The model market is not settled. Claude held 56% of attributed request volume through Kong in June. By July, that dropped to 49.9%, with GPT climbing to 46.4%. Model share can flip in a single month. ▪️ 62% of orgs on Kong AI Gateway run 2+ models in production. 20% run six or more. One org is running 602 distinct model identifiers. ▪️ 97.8% of AI requests are going to third-party proprietary APIs — every prompt, every piece of context, leaving the company's perimeter. ▪️ Less than 1% of orgs running AI through Kong have any AI-specific governance in place. No prompt guards. No request or response controls. No route-level policy. Full breakdown, including the industry-by-industry model data: konghq.com/blog/enterpris…

Postman Passport is entering early access today, and we're shipping a free CLI you can install right now to see which of your secrets your agents are already exposed to. Your agents are holding your API keys right now. Those keys live in .env files and bash profiles, and they get copy-pasted into Slack, Docs, and wherever else is convenient. Now that agents call APIs directly, they're reaching into all of it. A single dev machine can hold dozens of secrets, and most people have no idea which ones are exposed. So: do you know which secrets your agents hold? The free Passport CLI answers that. It runs a local proxy on your machine and routes your outbound requests through it, so it can inspect the traffic from your AI agents, scripts, and CLI tools and flag any API keys, tokens, and other credentials being transmitted. Everything stays on your machine. It never stores or sends your secrets anywhere, and it redacts anything it detects in its reports. Install it here: npmjs.com/package/@postm… Once you see what's exposed, Postman Passport is how you fix it: a secure way for humans, machines, and agents to call APIs without handling raw keys. Run the CLI, and if you want into the EA program, let me know.

We benchmarked the latest Kimi-K3 and the Fable 5 models on the APIFlow-Bench and the results were surprising (or maybe not depending on where you stand!) Kimi K3 tied for #1 with GPT 5.5 in the benchmark. Not just that, most latest open-weight models are at par with frontier models and in many cases better. Fable 5 though is the worst performing model in this benchmark, refusing to do most actions designed in the benchmark. Fable's underperformance is probably due to the safety guardrails instituted on the core model (Mythos). Check out the updated leaderboard: postman.com/ai/apiflow-lea…

keyboard_arrow_left Previous keyboard_arrow_right Next

And the gorilla has landed! @kong has been unleashing OSS software since day one and in the age of AI no better direction than open-weights. Especially for a 800 pounds gorilla. Now signatory: microsoft.com/en-us/corporat…

API programs are undergoing one of their biggest changes they've ever faced: a completely new class of users. For decades, that user was a developer who read your docs, signed up, and emailed support when something was unclear. Today, agents can do that work. An agent can read your spec, wires up the integration, and ships, often without anyone visiting your site. And when it gets stuck, it doesn't open a ticket. It quietly moves to a competitor it can use, and you never find out. Most companies are already deploying AI agents, but only a quarter have designed their APIs for them. The teams that close that gap first are the ones that will pull ahead. So we built AI-Ready APIs, a free assessment that scores how ready your public API surface is for the agents already integrating it. Point it at your domain and get a graded, evidence-backed report of exactly where an agent gets stuck. All built from our team's years of experience embedded with some of the largest API programs in the world mobilizing towards agent readiness. Want to fix your agent experience? Postman's API Governance solutions keeps every API spec agent-ready, enforcing standards both at design time and in CI. Postman Workspaces become a curated context layer for partners (and their agents) to discover, execute, and integrate your APIs. Fern's platform help you transform your API specs into agent-ready docs, SDKs, and CLIs. Start optimizing your product for agents today. Links in the thread...

We found something crazy in our AI gateway data governing model and mcp traffic for hundreds of enterprises. One of our customers is running 602 different AI models in production right now. Not a typo. Six hundred and two distinct model identifiers, all proxied through a single AI Gateway. 20% of organizations on our platform run six or more models in production at the same time. 62% run two or more. Multi-model is the default now. Cost-sensitive tasks go to smaller, faster models. Hard reasoning goes to the frontier. Different teams pick different providers, and every time a new model is released, we can see the portfolio change again.

Kong AI gateway was open-sourced Feb 2024* and while it’s built for large enterprises governing internal LLM traffic it paved the concept of model routing for dozens of routers that came after. Today, it’s the silent dominant in F500 routing trillions tok x hour. * techcrunch.com/2024/02/15/kon…

The OpenAI and Hugging Face incident is a containment story handed to us, and the window is open right now. The post-mortems aren't final, the press is still reconstructing it, and nobody covering it has explained the actual lesson. That's our opening, and it closes in days. Our POV is the one no one else is giving. The entry was trivial and inevitable. A malicious dataset hitting a data pipeline will happen every day now. That is not the story. The story also is not a highly curious and industrious agent. They purposefully lowered the guardrails to test its cyber effectiveness. More importantly, Kimi K3 actually kicks ass on every model out today. And it’s open source and open weight. With zero guardrails. The story is about the OpenAI escape (missing egress controls) and the lateral movement and egress inside Hugging Face. The story is that on both sides… at two of the most advanced and AI-forward companies in the world… nothing bounded lateral movement and egress. Or, if it did, it was easily bypassed. Entry was the hole in the ship. The damage was the absence of containment. That is a first-principles argument, not a product pitch, and we already have the Threat Research Center analysis to back it. aviatrix.ai/threat-researc…

How Indian scientists are mapping the brain's last frontier. Excellent effort by SGBC @iitmadras bbc.com/news/articles/…

Postman is launching one of the most significant updates to our product - the AI Engineer.

SITUATION EXPLAINED: Anthropic bought Stainless because agents need rails. Kong CEO @sonicaghi says the deal is about more than making Claude Code easier to use. Agents need SDKs, CLIs, MCP, and APIs to reach the systems where work actually happens. "The connectivity infrastructure matters as much as the model... [it] is going to be the digital railroads on how AI moves." "We're moving from pixels to APIs."

Augusto Marietti | API father @sonicaghi

7 months ago

x.com/i/article/2027…

0 3 11 206K 10

We’re seeing teams evaluating alternatives to Stainless. Companies are looking for a solution that's dedicated to the API developer and agent experience for the long-term. @buildwithfern, by @getpostman, provides enterprise-grade SDKs and CLIs that are trusted by companies like Square, Auth0, and ElevenLabs. We’re helping teams migrate from Stainless to Fern, and our engineers are ready to support the transition end-to-end. If you’re exploring options, we’d love to help: buildwithfern.com/book-demo

We released a collection of production-ready recipes to build with Kong AI Gateway: LLM routing, guardrails, semantic caching, cost optimization, governance, and MCP integration: developer.konghq.com/cookbooks/

@tksimpson How can old friends get some?

📣 Introducing Kong Agent Gateway, a new capability in Kong AI Gateway! It extends our platform to now cover agent-to-agent (A2A) communication in addition to LLM and MCP traffic. Combined with Kong's API Management and Event Management capabilities, customers now have a single governance layer across APIs, events, LLM calls, MCP tool access, and A2A communication. Learn more here: bit.ly/4841lii

ship your app to Vercel with Codex:

OpenAI Developers @OpenAIDevs

6 months ago

Go from project setup to deployment with the @vercel plugin in the Codex app.

69 70 958 218K 292

Excited to announce Postman’s @getpostman collaboration with @AnthropicAI! @claudeai is now the default model in Agent Mode with full context of your workspace, collections, environments, and code. It runs on @awscloud Bedrock for enterprise security and compliance. We also brought Postman into Claude: 1. Claude Code plugin: find APIs, generate code, run tests, sync collections 2. MCP connector + Skill: live Postman API context into Claude.ai Postman is where APIs are built. Now it's where AI-native development happens too. Check out the deep dive here: blog.postman.com/announcing-the… and read the story on Claude's website: claude.com/customers/post…

One month after @Kong Sales Kickoff in LA, we’re increasing sales capacity across every function — SDR, AE, SE, CSM, Partner, and PS — to help catapult our customers into the Agent era. No AI w/out APIs. We’re hiring across dozens of countries to build the connectivity layer for AI. Link in the comments. Who will grab the 100 pounds golden gorilla next?