Kate Huggins

Interested in changing my own mind not yours. Tweets = notes to self.

[Kate Huggins @Huggette]

💀

[suz @suzwhoever]

shoulda known Ed Sheeran was evil when I learned his wife worked at Deloitte. Imagine having an insanely rich husband and choosing to go to work at Deloitte every day

[Kate Huggins @Huggette]

@FraserNelson Australia’s immigrant students outperform to an even stronger degree

[Kate Huggins @Huggette]

The clean up the next day was mammoth. He did most of it but I took on the (almost destroyed) carpet. This is really to say: let them throw the party, let them feel the feels, let them make the mess (and clear it up).

[Kate Huggins @Huggette]

I was in my bedroom listening to 65 kids really feeling “so close to you right now”, and I was just so happy he’d been able to create a little irl force field for them to feel jubilantly, collectively, alive in a world where that is harder than ever

[Kate Huggins @Huggette]

@JayHowl2 @megievalist Gosh this is beautiful

[Kate Huggins @Huggette]

@megievalist

[Kate Huggins @Huggette]

@rjmontgomery $14k for maintenance paint job on timber cladding (3yrs old) covering approximately 20% of my 4 bed house.

[Kate Huggins @Huggette]

@helenlewis Shocked by the lax regulation (age and number of children) and 1-sided vetting (criminal history). Interesting the practice has shifted so significantly to donor / IP eggs. Thanks for the gift read.

[Kate Huggins @Huggette]

Kinda wild

[Peter Schiff @PeterSchiff]

The nickel and copper in modern nickels are now worth 7.76 cents, 55% more than the cost of buying nickels from the bank at face value. The U.S. government stopped making pennies last year. Soon it may stop making nickels too. Buy yours while you can. Much better than Treasuries.

[Kate Huggins @Huggette]

@JosephNWalker @Charlton_AB About as good an interview as you’re going to get from a politician.

[Kate Huggins @Huggette]

@janet_e_egan Excellent questions, even better answers . Keep doing podcasts please.

[Neil Renic @NC_Renic]

They're calling it the most relatable diary entry of all time

[Kate Huggins @Huggette]

@NiallHarbison You are helping both those suffering with depression and those that live with them by sharing this way. Perhaps more important than the many dogs you save and care for. Perhaps the dogs are here to help you so you can help people like you. I hope the weight lifts soon.

[Kate Huggins @Huggette]

@DrCameronMurray Helpful. We need to lift eligibility barriers: higher language standards for student visas, higher income thresholds for temporary skilled workers, tighter limits on secondary (family) visas, and stricter rules for bridging visas that allow visa-swapping and extended stays

[Kate Huggins @Huggette]

@michaelchen Weirdly effective

[Arram @arram]

Reading the METR Hugging Face Report is giving me a taste of the post singularity world. Thousands of agents thinking full time at AI speeds means too much happened for investigators to confidently understand it in detail.

Soon the entire world will be moving too quickly for us to understand. Prepare to live in bewilderment.

https://t.co/CLEJ5Wo4OH

[Kate Huggins @Huggette]

Need

[wast3 @0xWast3]

Anthropic ex-engineer runs an internal graph that costs $6 a month and catches what a $300,000 eval suite misses. No retrieval layer. Seven nodes and one rule about who's allowed to change their mind.

He published the whole schema.

His version starts from the opposite idea. A graph is not an execution order. It's a memory of why.

Seven nodes. Every edge carries the reason it exists:

> INTENT - states what the task is for. Never how
> DECOMPOSE - splits it into steps, each with a stated assumption
> WORKER - executes one step. Sees nothing else
> AUDIT - checks the output against the assumption, not the goal
> DRIFT - compares the current step to INTENT and flags divergence
> LEDGER - stores every decision with the assumption that justified it
> ROOT - holds the graph, and when an assumption breaks, re-runs every step built on it

Six nodes act. One node remembers why they acted. Every step carries the assumption that made it correct, so a false one only reruns what stood on it.

That's the entire design. A pipeline that forgets its reasons has to redo all of it or trust all of it.

He replayed a month of agent runs. 4,100 steps, 380 built on an assumption that was wrong by day three. The old pipeline shipped all 380 and linked none of them.

Everyone else builds graphs where output moves forward and the reasoning evaporates. He built one where the reason travels with the result. The blast radius is the output nobody else produces.

The article below is the full build - node prompts, the assumption format, the invalidation rule that finds every step downstream of a broken one.

Save it. You'll want it open in the other

[Jacob Matson @matsonj]

@SamBroner I think its tough - if you only have MCP you lose visibility to how users are using it (i.e. traces). I think what you probably want is an agent for your normie users, and then MCP for power users.

this mismatch of course here is that you make more revenue on casual users