Andrew Freedman, CFA 🦅

I am going to be in San Francisco for meetings 3/24 - 3/25… ping Kevin/Josh or sales@hedgeye.com if you would like to meet!

@LOGOinvestor lol 😘

@HedgeyeComm Nobody loves cryptic tweets like you baby!!

The gap is about to get really wide…

Meta just confirmed 16,000 layoffs but I'm hearing the real number is closer to 22,000 Sources inside are saying this isn't just about AI efficiency gains - it's full workforce replacement One senior engineer told me they've been documenting every decision tree, every debugging process, every architectural choice into "knowledge management systems" for the past 8 months Turns out those systems were feeding training data to internal AI agents The AI can now handle 73% of the codebase modifications that used to require human engineers Middle management got the directive last week: identify which engineers are "AI-multipliers" and which are "legacy dependencies" Word is they're keeping 1 senior engineer per team and replacing the rest with a combination of Claude API calls and offshore contractors running Cursor The offshore team costs $28k per engineer annually versus $240k for Bay Area talent Performance reviews next month aren't actually reviews - they're knowledge extraction sessions Engineers think they're mentoring junior teammates and training the next generation They're actually teaching the AI systems their last remaining institutional knowledge One source showed me the internal productivity metrics: teams with heavy AI integration are shipping 340% faster with 60% fewer engineers The scariest part? The engineers being laid off are the ones who built the AI tools in the first place They spent two years automating themselves out of existence and calling it innovation Badge deactivations start Monday But sure, keep telling people Meta is "investing in the future of work"

Hard pivot to become Anthropic

Introducing Dasher Tasks Dashers can now get paid to do general tasks. We think this will be huge for building the frontier of physical intelligence. Look forward to seeing where this goes!

!!!! #AgentLayer

MCP not dead…. It still very much so has a purpose when it comes to discovery and non-deterministic functions. People just waking up that you can put a LLM wrapper around a CLI and click go… which is what Claude Code is btw… The people who are saying MCP is dead in the tech community are the ones who were trying to make MCP servers do things it was not designed to do… Pipes aren’t shifting… just look at what Google is doing with UCP and WebMCP… you don’t rewrite the entire internet overnight.

@HedgeyeComm Agree on both points, but think mainstream adoption will take a while. The "pipes" for agentic keep shifting. First MCP was it, now MCP is dead. These shifts, including transacting on mobile, take longer than most think...

I think 24 months is the likely timeframe. V1, was DOA...

@buccocapital Kidding!

“It organizes your files” “It prioritizes your emails” “It tells you insights about your calendar” These are not real things. They are not making you more productive. It is making you an idiot Yes, AI is great. But this is fake productivity. This is dumb. You are being dumb

@WaterworldCapi1 $TSM $TSM

Deep thoughts

@sama Sam’s eulogy for software engineers

@HedgeyeComm amazing call

Tomorrow we will unveil the all new vibe coding experience in @GoogleAIStudio, the team has spent 4 months rebuilding it all from scratch and smoothing out rough edges to help everyone bring their ideas to life. This is a big step forward, but just the start : )

😜😜😜

Ultimately it comes down to the contract definitions which none of us know. I understand your point re stateless API being used to build stateful agents. Where state is managed not at the model level but in the harness. My point still stands though… if OAI ends up building “models” that are actually agents, while technically that requires a stateless API call - does wrapping it and shipping it as a stateful endpoint a technical loop hole? Also, what happens if OAI ships a version of the model with LTSM? Or embedded RAG? Is that technically stateful?

@HedgeyeComm I don't think this means what you think it means. Microsoft simply has to write their own agentic systems that integrate with the API. Opencode, Openclaw do this btw. Stateless in this case just means that the state is contained with the agent harness and passed to the API

We were always heading down a path where the models were going to become agent systems themselves... Which is potentially a big problem for $MSFT because their deal with OAI/Azure is for stateless API... AND if the rumors re 5.4 are true and it is agentic and stateful...