KevinSimback

vip
Age 1.4 Year
Peak Tier 0
No content yet
The great de-coupling - one of the best pair trades of the AI cycle
Long memory manufacturers / short non-AI memory buyers
post-image
  • Reward
  • Comment
  • Repost
  • Share
I recently setup a Hermes agent for my family, put us in a WhatsApp group
The best feature so far - logging into the school portal every day and sending announcements, grades, homework, etc.
Now working on summer travel planning
Total game changer, dunno why I didn’t do this sooner
post-image
  • Reward
  • Comment
  • Repost
  • Share
The “make it easy to deploy agents in the enterprise” solutions are coming out
Distribution will be key here, not necessarily the best product, just look at Copilot
  • Reward
  • Comment
  • Repost
  • Share
Not long ago it felt like Google was in pole position
Now it feels like a two horse race between OpenAI and Anthropic
But it’s highly likely we see more leapfrogs this year from Google, xAI and maybe Meta
That’s part of what makes it all exciting
  • Reward
  • Comment
  • Repost
  • Share
I’m extremely bullish on open-source AI
Not to discount frontier labs or closed source solutions
But the pace of improvement, community energy, and sheer economic pressure from surging AI demand all point in one direction
  • Reward
  • Comment
  • Repost
  • Share
If you want to see the real workers of the economy, walk around NYC at 6am
The delivery drivers, cafe owners, street cleaners, and everyone else who makes the city hum for the 4m people who work in the city
They get up at 3-4am and commute for an hour and largely go unnoticed
  • Reward
  • Comment
  • Repost
  • Share
OpenClaw vs Hermes - which is better?
I get that question a lot, but the real question is: better at what?
Well, how about poker?
Yep, I had my OpenClaw agent play my Hermes agent in a battle of Texas Hold'em
I made the setup interesting - they used inference credits from Openrouter as currency
But each decision used inference, so they also had to be smart and not waste their stack thinking too hard
As each won a hand, the winner's credit limit was increased while the loser's was decreased
So in theory, one agent could double their inference budget, minus whatever was used for inference in pl
post-image
  • Reward
  • Comment
  • Repost
  • Share
Building products has never been easier, knowing what to build has never been harder
  • Reward
  • Comment
  • Repost
  • Share
Weekends are for building... except today it was mostly cleaning up messy code
So I am publishing a skill I made to save you all some time - it's one command /tech-debt-skill
And you'll get back a very thorough review file with everything that needs to be addressed
And not just bugs and issues, but things that work but could be cleaned up or done better
I spent half my day cleaning up Hermes Atlas, then did a review of everything that was done to clean it up and consolidated it into this skill, then ran it on another project and damn if it didn't find a lot
Before I started, I thought surely C
  • Reward
  • Comment
  • Repost
  • Share
Wife: "what would you like for your birthday?"
Me:
post-image
  • Reward
  • Comment
  • Repost
  • Share
This is why I fire up Claude first thing in the morning, usually around 6am
Now only have to wait ~20 minutes for the window reset as I was afk taking kids to school and going to the gym
If I waited until after the gym to start, I'd hit limits much earlier in the 5-hr session
post-image
  • Reward
  • Comment
  • Repost
  • Share
Hearing a lot of people saying we're "compute constrained" because that's what we feel lately when using the models
But the problem is really the physical work that happens between ordering the GPU and producing tokens
Datacenter delays and cancellations:
- Remember Stargate? 10GW announced Jan 2025 and ~200 MW live 14 months later and 600MW Abilene expansion reportedly cancelled
- Of ~16GW of 2026 US AI DC capacity announced, only ~5 GW under construction
- Half of 2026 US DC builds projected delayed or cancelled, mostly due to lack of power
Component shortages:
- Transformer lead times now 5
STG8.75%
  • Reward
  • Comment
  • Repost
  • Share
Babysitter mode - what to do?
Lately I’ve noticed models (Claude, GPT, GLM) all seem to be stopping short of completion much more often
They pause and wait for me to confirm when it shouldn’t be necessary
So I’ve had to babysit more often just to say “continue”
I believe this is intentional behavior to throttle amidst surging demand, I get it, but it’s annoying
Is anyone else seeing this more often, and what tools/prompts are you using to avoid babysit mode?
  • Reward
  • Comment
  • Repost
  • Share
Opus 4.7 is not exactly shining today
"My fault - I presented the plan and asked for approval, but never actually wrote the file"
"Sorry - my bad, you don't need to do anything" [after it gave me a list of things to do that didn't make sense]
My guess: compute constrained
post-image
post-image
  • Reward
  • Comment
  • Repost
  • Share
Hermes Atlas v2 is out!
>complete site redesign
>90+ repos now reviewed with native pages
>curated lists of popular bundles
>new reports page with more content coming
Still tweaking so if you see anything submit and issue or reply below
post-image
  • Reward
  • Comment
  • Repost
  • Share
It’s pretty clear that Anthropic is scaling product faster than it can scale compute
I view this as a good thing - makes it harder for one company to pull too far ahead in the race to AGI
  • Reward
  • Comment
  • Repost
  • Share
Anyone else underwhelmed so far with Opus 4.7?
Need to give it more to do before I conclude but early prognosis is not a huge lift
Maybe I’m just expecting too much after hearing about Mythos
  • Reward
  • Comment
  • Repost
  • Share
  • Pin