Michael Levan👨🏻‍💻☕️

An SRE Agent needs 4 things: 1. A proper prompt/system message. This will give the direction for the Agent to take. 2. Agent Skills that are designed to turn your Agent into an SRE and observability expert. 3. A good LLM 4. Targets Number 4 is where the "magic happens". When you put together a solid Agent for a particular use case, the target can be anything in its speciality. For example, in an SRE Agent, the target could be anything from a cloud provider to a Kubernetes cluster. #agenticai #ai

💰 Progressive Disclosure for MCP: 𝙨𝒂𝙫𝒆 𝙩𝒐𝙠𝒆𝙣𝒔 I was able to reduce token usage by 9͟1͟.͟1͟%͟ with agentgateway. The verified token results: - With Progressive Disclosure: `usage.prompt_tokens` = 970 - Without Progressive Disclosure: `usage.prompt_tokens` = 10877 The test measures prompt-token overhead. Each test does an MCP handshake against `/mcp`, fetches `tools/list`, converts it into an OpenAI-style tools array, then sends a single Anthropic request to `/anthropic` with that tools array inlined in the request body and `reads usage.prompt_tokens` from the response. #agenticai #ai

But... you're not denying Pro subscribers won't be impacted once the "test" is done, lol.

When you’re building Agents, they either need: 1. Dedicated permissions 2. Permissions on behalf of a user Or in some cases, both. Dedicated permissions means that every single Agent has its own AuthZ. This can get cumbersome for large orgs, so the Agents can act on behalf of a user (OBO). With an implementation like OBO, an Agent has the same permissions as the user that is using the Agent. #agenticai #ai

“40 Agents and we don’t know what half of them do” When your workflow consists of opening Claude Code on your terminal, that’s one thing… But when you have multiple Agents working autonomously? Shadow AI is becoming a real big deal for organizations and when Agents are running and no one knows what they do, that attack surface is just begging to be exploited. Auto Agent discovery across your environment is key. #agenticai #ai

@nnennahacks jelly

Made it to Miami! Sooo nice to be in 80 degree weather 😎

Statefulness is crucial for Agents. Imagine you have an autonomous Agent performing a task and for whatever reason, the Agent goes down, you try to resume it, and there's just... nothing. That's going to be the "I accidentally put my tape drives on the wall with an X-Ray machine behind it" moment. #agenticai #ai

@virtualized6ix "just to wet the whistle"

This thing is so cute at making coffee

Do me a favor folks. Build your projects and share them. Who cares if they get zero GitHub stars or you never touch them again. Build. Share. Community. Learn It’s what we’re all here for. Don’t listen to the gatekeepers that say no to build.

Why would Opus 4.7 think my Agent Skills config is malware? Lol

@AnthropicAI Claude 4.7 is off to a great start #agenticai #ai

This Week In AI breaks down: - Security before Agents - The architectural pattern of AI implementation - Agent Sandboxes Link below. michaellevan.substack.com/p/a-week-in-ai… #agentai #ai

Opus 4.7 honest review? It did what Opus couldn’t do for the past 5 weeks. Figure out a solution better than GPT5.4 What have your thoughts been on it? #agenticai #ai

@SebbyCorp @hashishrajan

@hashishrajan I am ai. Ai is me and I have replaced me with ai

Which stage of AI adoption are you actually at right now?

When you put yourself out there, you will fail 99 out of 100 times. The question is - do you want to keep trying or would you prefer to stay safe?

@SebbyCorp @John_Capobianco John, you in Toronto today?

@TheNJDevOpsGuy @John_Capobianco is at toronto nog

Any tech meetups happening in the Toronto area today?

I’m seeing a lot of “Anthropic is taking everything and X company is doomed”. This is really silly. When has there been one company that was the only one doing “the thing”? In the early days of computing? Do you know how many companies do the same thing as 20 others? This isn’t new… they’re just breaking into the market.

In the past two months I’ve: 1. Traveled to 3 countries 2. Traveled multiple states 3. Talked to hundreds of people 4. Worked in incredibly large environments helping folks implement AI I love my job. @soloio_inc