Marco Palladino retweeted
Excited for Kong to partner with @ServiceNow to help enforce Agentic traffic from IT to Infrastructure. newsroom.servicenow.com/pres…
And this doesn’t count all the MCP traffic that’s been consumed together with those tokens
500T tokens. That’s how much traffic @sonicaghi says is moving through @Kong’s gateways today.
Run the math on 140x YoY growth and the volume roughly doubles every 7 weeks. At that rate, a quadrillion is a near-term milestone.
Cheaper tokens won’t cap that. They’ll spread it across more models. Enterprises asked Kong how to secure, govern, and cost-control API calls a decade ago. Now they’re asking about tokens.
His read: “In a non-deterministic software enterprise world, the need for deterministic AI gateways will just increase.”
Had a blast talking at @MTSlive about the next things to do in the agentic transformation to fully automate our organizations! And of course, to build agents with volcano.dev 🌋
Watch at youtube.com/watch?v=OXa_n3uu…
Marco Palladino retweeted
Kong + Headroom
We are proud to announce that Headroom Labs is partnering with @kong
@kong released their new AI Gateway release 2.2, which has a native integration with Headroom. When Headroom was first announced, @sonicaghi commented on our post, that he'd like to integrate it into Kong, and we have spent the past few weeks making that comment a reality :)
We are very thankful to @sonicaghi and the Kong team - for being vociferous advocates of token intelligence.
Full release blog: konghq.com/blog/product-rele…
So many announcements today ❤️
Now GA: Kong AI Gateway 2.2, featuring a new AI-native user experience, advanced cost management, and broader model support.
🔌 MCP Server Bundling: consolidates a growing sprawl of MCP servers into one Kong route, revealing only the tools each caller is authorized to see and execute.
💰 Modality-aware cost governance: tracks text, audio, image, and video pricing separately, plus cache reads vs. cache writes, for accurate cost attribution instead of flat token estimates. Also ships support for Headroom, for advanced prompt compression.
🪪 Identity-aware AI policies: powered by Kong Identity, so rate limiting, analytics, cost attribution, and access control key on an authenticated principal instead of a gateway-local Consumer.
☁️ Native AWS IAM authentication for Amazon Bedrock AgentCore: declarative, platform-managed SigV4 auth, no more hand-rolled workarounds.
🌐 Expanded model and provider coverage: native support added for Kimi, Microsoft Foundry, and Amazon SageMaker.
🧩 Support for Jev: adopt this emerging decision-oriented model without building a separate integration and governance path.
Learn more: konghq.com/blog/product-rele…
I have been a little quiet on here recently because I was heads down working on a new product that I consider the most exciting product I have ever released. And today at Kong AI Summit we finally announced it.
Introducing 🌋 Volcano. It provides the AI stack for building AI agents at scale:
- Native Claude Code, Codex and Cursor plugins
- Durable workflows + durable tracing
- Branchable Postgres database + file storage
- Unlimited user authentication + realtime capabilities
- Distributed locking to coordinate agentic work across crews
- Full blown Next.js hosting behind our CDN across multiple regions
- Distributed backend hosting for APIs and MCP tools
- Sandboxes GA in a month
- Javascript, Typescript, Python and Ruby support
You can use Volcano standalone, and you can also use it in combination with our Kong Konnect platform to govern, secure and discover AI agents (and APIs, MCP tools, and LLMs) at scale.
You can get started for free at volcano.dev - I will be posting a lot about Volcano in the coming weeks as I am 100% obsessed by this product (..and I hope you will too).
Crazy thought but...The first cities on Mars may not have farms. They could have food factories. No soil. No livestock. No fields. We already have much of the technology.
Martian CO₂ + water ice + minerals from rocks + solar/nuclear energy → protein, fats and carbs.
On Mars, agriculture may look more like a data center than a farm.
🤖 Made with AI
The Nazis once made a butter-like substance from coal in a plant located in Witten. It worked but took approximately 60–70 kg of coal to make 1 kg of synthetic fat.
The concept would be similar: Solar power + CO₂ + water + rocks → food.
Modern bioreactors could potentially convert something like ~3–10% of sunlight into food energy, several times the efficiency of crops.
Who would have thought that we would have had to create AGI applied to advanced robotics to eventually fix the huge problem that is folding clothes.
Nothing else worked.
Marco Palladino retweeted
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/ko…
Marco Palladino retweeted
SITUATION EXPLAINED: Have the human internet and the agent internet already split apart?
We asked @subnetmarco, CTO of @kong
"There is already a bifurcation. If we look at our properties right now, 80% of traffic is being driven by agents. It's not even being driven by humans anymore."
"Not being able to properly cater to this new agentic traffic means that we're not catering to our users anymore, our customers anymore."
"This is not something new. If you were not on the Yellow Pages, your business didn't exist. Then everybody moved to the internet. Then mobile apps. Facebook acquired Instagram because they felt like they were not catering to a whole new buyer moving to mobile."
"Likewise, we're seeing that shift yet again, every time there is a shift in the distribution channel, that creates opportunity. It also creates disruption for the ones that cannot follow where the trend has gone."
"What determines if one succeeds or fails is the ability to cater to this new buyer, and that is the agent itself."
1. Agents are simply applications, containerize them and deploy them. Agent-per-container is a real thing. Scale horizontally as needed.
2. Delegating work across agents is like delegating jobs across a distributed environment: leader election, FIFO queues, and so on. We have been doing that for decades.
3. The best durable workflow is the one you store in Postgres. Use in-memory caches as needed for performance. Just like the past 20+ years.
4. Use your favorite LLM where decision requires uncertainty or judgement, use deterministic code for everything else.
I am at an Airbnb in Europe and turning this on should be the “Will Smith eating spaghetti” test for AI robotics. Once it happens, that’s how we know we invented AGI.
One of the best ways to avoid model lock-in is to use a multi-LLM AI Gateway that allows to seamlessly send requests to multiple models without having to rewrite your code. At @kong we built an enterprise solution that allows to achieve just that, via the universal API. Build once, switch to any model instantaneously.
At this point every CEO should be asking what their strategy is to avoid model lock-in.
If it isn’t clear what Anthropic is doing, it is:
- build something amazing
- decide who gets to use it after you prompt it if the prompt falls into areas they deem unacceptable by their sole standard
To be clear this is completely above board and legal. It’s just an idiotic risk for corporate users to bear especially as the coding models become equivalent.
The business continuity risk will become more obvious as companies accidentally trip over Anthropic’s ToS and have to decide if they will subsume their business viability to them by doubling down on Anthropic models or find open source (and, btw, much cheaper) alternatives where they are in control.
As stated previously, get ready to be inundated with the term “control plane” which is the natural solution to this problem.
Shameless plug - this is what 8090’s been building as we expected this moment to arrive…
If you’d like to learn more: 8090.ai
