Ever since I stopped using my anon account on X and started using this one, my life has improved a lot.
Zero politics. Zero conspiracies. Zero doom.
If I see any of it, I SLAM not interested.
This is the experience I always wanted.
I just hope I can keep it this way.
Recuris is a commit problem, not an RSI pitch.
Working memory tracks what is still open. Skills get pulled from that state, not from the full chat.
The model stays frozen. A Meta-Agent patches only the memory piece the structured trace blames. A gate has to keep the patch.
35 of 37 model and benchmark pairs went up.
On tau2-Retail, GPT-5.6 Sol went from 58.3 to 76.1. Claude Opus 5 went from 72.4 to 87.9.
A structured trace localized injected faults at 64.8. The outcome bit hit 13.0.
Those lifts still get me. Memory is not a solved problem.
https://t.co/iKpeEnBIQa
So what is this?
A MCP? Exclusively supports only claude?
New product from SF?
This turned me away because it was all around claude
Good thing we’ve been building variations of this internally. Helps when all data across the org is available from snowflake.
Hermes real-profile browsing is a copy of my active Chrome logins. Not my live window.
Off by default. Hermes snapshots cookies and logins, then drives its own Chromium.
Settings, Browser, Use My Real Browser Profile. Or set browser.use_real_profile true.
I would leave it off until the agent needs a site I am already signed into.
It does not solve captchas. Cloud browsers still own that.
Turn it off and the snapshot is deleted on the next browser use.
https://t.co/BPcgypLsGq
Hermes Agent can now seamlessly browse as you.
Turn on real-profile browsing and your agent acts with your logins, from a managed copy of your existing Chrome profile.
🚨 This is a BIG upgrade for anyone using Hermes for data work.
Hermes can now run execute_code with a SESSION-PERSISTENT Python kernel. 🔥
Load a 2GB dataset once.
Build your DataFrames.
Run analysis.
Ask another question.
Keep going.
Hermes doesn’t have to rebuild the entire Python environment every damn call anymore.
That’s a proper agent workspace. 🤘
Most agent memory benches score recall.
StateMemBench scores whether the answer uses the current fact or a superseded one.
234 multi-session scenarios. 322 graded probes. Closed-pool grading for state drift.
StateMem lifts current-state accuracy about 1.8x over the strongest same-backbone memory baseline on DeepSeek-V4-Flash (0.199 to 0.363).
https://t.co/h21nV2wJ6J
We gave our AI system a spec, and in under 2 weeks, it designed, verified, and deployed a chip that beats NVIDIA.
It’s built for low-power physical AI workloads. We’re running live inference on >B+ parameter models like Llama, Qwen, and Kimi, serving at 3.4x better perf/watt than NVIDIA Jetson.
From just a specification, our AI system autonomously generated all of the RTL Design, UVM verification, formal proofs, firmware, drivers, and kernels, co-designing the model, software and silicon as one optimization loop.
Better AI can now design better chips to run AI, leading to a loop of recursive-self improvement towards our path to abundant intelligence.
While America begins to rightfully castrate Canada economically, please consider the Americans who live here. Many of us have been here for years, long before the tensions between our governments.
The people will suffer through this.
Penance for the way the majority of us vote.
✌🏻
I'm really excited about this.
As an American living in Canada for the last 15+ years, I see this as potentially the way I can get out of here and back to the US.
Can't wait for more crypto adoption.
My BTC is becoming one of my exit strategies from Canada.
Crypto-backed mortgages have moved in.
Borrowers in the US can now use Bitcoin as collateral for a down payment - without having to sell it or face margin calls.
Plus, Coinbase One members can get up to $10,000 back at closing.
@tonysimons_ No joke my first project when I started using hermes back in like January or December was something called “Fleet” and I used it to manage a fleet of hermes agents deployed across their own containers
I feel like shortly after I see this word all the time now “fleet”
So @cursor_ai, the additional token rate charge of $0.25 per 1 million tokens needs to be scrapped. I have to pivot to new tools because it’s clear an effort is underway to phase out Cursor. Not happy about it. This single CTR charge is so damaging. Make your money on VTDs, not up-charging clients.
@Da7_Tech Data privacy and ownership are something I never talk about anymore.
Those who know, know, and everyone else is a free product.
People are clueless about the amount of exploitation that takes place with the data they volunteer daily