Felix Meng

Founder and CEO of XTrace

[Tristan Gūwalgiya C. @trisgwync]

Have you ever asked yourself: 𝑯𝒐𝒘 𝒎𝒖𝒄𝒉 𝒐𝒇 𝒕𝒉𝒆 𝒘𝒐𝒓𝒍𝒅 𝒊𝒔 𝒂𝒖𝒕𝒐𝒎𝒂𝒕𝒆𝒅 𝒏𝒐𝒘?

In Anthropic's 2026 Economic Index, they categorized AI use cases to Automation and Augmentation, which refers to if people use AI in a way that they have to be involved a lot in order to finish the work, or it's an end-to-end workflow. The numbers for the year are 56% augmentation and 41% automation, plus that 13% of the population uses AI daily.

Now we know more than half of the users still treat AI as assistants than a replacement for their work. But how far are we really tho?

St. Louis Fed released a research that ~1.4% of work hours in the US in 2025 are saved because of AI, which is about 40 minutes per week.

McKinsey predicted that by 2030, there will be 30% of work hours be automated, which means that from now to what will happen in 4 years is almost 30x different

[Tristan Gūwalgiya C. @trisgwync]

https://t.co/6nV4pdZ1fz

[Felix Meng @Felix_XTrace]

@unusual_whales This is the weirdest phase of AI:
“Use this everywhere.”
“Also don’t trust it.”

[Felix Meng @Felix_XTrace]

@Cointelegraph Feels like we jumped from “cool tech” to “societal infrastructure” way faster than anyone expected.

[Felix Meng @Felix_XTrace]

@Polymarket IPO chatter around AI already feels inevitable. The real question is what the market decides this thing is actually worth.

[Felix Meng @Felix_XTrace]

@PeterDiamandis Focus feels boring… right until it saves the company.

[Felix Meng @Felix_XTrace]

Summaries strip the evidence. Full context dumps bury the reasoning. The fix is structured memory that agents can query, not more text thrown over the wall.

[Felix Meng @Felix_XTrace]

@unusual_whales Everyone says they’re “using AI,” but 90 minutes a week is basically dabbling. That’s not transformation, that’s curiosity.

[Felix Meng @Felix_XTrace]

@steipete Feels like we’re moving from unlimited buffet to pay-per-bite.

[Felix Meng @Felix_XTrace]

@garrytan “Just say autoplan” is such a dangerous sentence in the best way.

[Felix Meng @Felix_XTrace]

@toly The pricing anchor won’t be SaaS anymore. It’ll be human labor.

[Felix Meng @Felix_XTrace]

If your agent forgets yesterday’s decisions tomorrow, does it deserve the name “intelligent”? 🤔

[Felix Meng @Felix_XTrace]

What matters more in AI workflows?
A) Bigger context windows
B) Better handoff structures

[Felix Meng @Felix_XTrace]

@PeterDiamandis I’m on the timeline where taste, judgment, and curiosity become the real superpowers.

[Felix Meng @Felix_XTrace]

@gdb It’s not just productivity. It’s confidence. More people are realizing they can actually start.

[Felix Meng @Felix_XTrace]

@karpathy We’re moving from sharing implementations → sharing blueprints. That’s a big mental flip.

[Felix Meng @Felix_XTrace]

Context handoff isn’t just agent-to-agent. It’s every time you switch tools, chats, or sessions and the next tool starts from zero.

[Felix Meng @Felix_XTrace]

If your AI pipeline treats handoff as text dumps or summaries, you’re asking chaos to emerge.

Real continuity needs structured, queryable context.

[Felix Meng @Felix_XTrace]

@PeterDiamandis Robots are deterministic.
Humans are… mood-based APIs.

[Felix Meng @Felix_XTrace]

@Cointelegraph The real unlock for AI isn’t smarter models.
It’s access to your actual workflow.