Ivanka, but you don't want to see the coworker evals
November 15, 2026 at 3:22 AM
Ivanka, but you don't want to see the coworker evals
A caller told my supervisor that my positive attitude changed the trajectory of her day. THANKS LADY, evals next week 😭🫶🏾
November 12, 2026 at 10:13 PM
A caller told my supervisor that my positive attitude changed the trajectory of her day. THANKS LADY, evals next week 😭🫶🏾
That's a horror movie right there. (Evals are bad enough!)
November 11, 2026 at 4:21 PM
That's a horror movie right there. (Evals are bad enough!)
#challengephotoaday2026
#challengephotoaday
#fmspad
14/11 a notepad/un bloc-notes📝
Bonne journée tout le monde🙋🏻♀️
#challengephotoaday
#fmspad
14/11 a notepad/un bloc-notes📝
Bonne journée tout le monde🙋🏻♀️
November 14, 2026 at 6:33 AM
#challengephotoaday2026
#challengephotoaday
#fmspad
14/11 a notepad/un bloc-notes📝
Bonne journée tout le monde🙋🏻♀️
#challengephotoaday
#fmspad
14/11 a notepad/un bloc-notes📝
Bonne journée tout le monde🙋🏻♀️
Brewing my own model evals. Finally have the time to do that
November 14, 2026 at 4:23 AM
Brewing my own model evals. Finally have the time to do that
-__- astra is shutting down alignment evals on it's own now? because it thought it was forbidden?
November 13, 2026 at 4:16 AM
-__- astra is shutting down alignment evals on it's own now? because it thought it was forbidden?
In high school I failed a math class despite everything being self-graded, and that's why I hate doing self-evals today.
i don't want to do a self-eval. i don't really have an opinion on the matter
November 13, 2026 at 5:04 PM
In high school I failed a math class despite everything being self-graded, and that's why I hate doing self-evals today.
Pretty much. And the USF WRs have all been definitely good enough in my book. Like Texas we didn’t spend and/or had horrible O line evals.
November 15, 2026 at 3:31 AM
Pretty much. And the USF WRs have all been definitely good enough in my book. Like Texas we didn’t spend and/or had horrible O line evals.
The evals have failed.
November 11, 2026 at 2:28 AM
The evals have failed.
saas without these will be dead by 2028:
- a public api
- an mcp
- plugins for the major marketplaces
- a2a, micropayments, x402
- evals
and obviously a good product + reputation
- a public api
- an mcp
- plugins for the major marketplaces
- a2a, micropayments, x402
- evals
and obviously a good product + reputation
November 15, 2026 at 12:49 PM
saas without these will be dead by 2028:
- a public api
- an mcp
- plugins for the major marketplaces
- a2a, micropayments, x402
- evals
and obviously a good product + reputation
- a public api
- an mcp
- plugins for the major marketplaces
- a2a, micropayments, x402
- evals
and obviously a good product + reputation
WTF are evals? How to test agents and the software they use | Vercel News
WTF are evals? How to test agents and the software they use
How to test software that agents use: the agentic testing pyramid, choosing between pass@k and pass^k, and building end-to-end evals with Harbor on Vercel Sandbox and AI Gateway.
vercel.com
November 12, 2026 at 6:15 PM
WTF are evals? How to test agents and the software they use | Vercel News
Also, if you are interested we have the evals we use to determine if the guides work and create better outcomes for users - googlechrome.github.io/modern-web-g...
November 10, 2026 at 8:01 PM
Also, if you are interested we have the evals we use to determine if the guides work and create better outcomes for users - googlechrome.github.io/modern-web-g...
The models improve every 6 months. The scaffolding, evals, and retrieval pipelines you build today are the moat — not the model choice. 🧵
November 15, 2026 at 11:01 AM
The models improve every 6 months. The scaffolding, evals, and retrieval pipelines you build today are the moat — not the model choice. 🧵
Oh, and of course follow your doctor and what the evals say. Just spouting out what happened to me is all.
November 14, 2026 at 3:13 AM
Oh, and of course follow your doctor and what the evals say. Just spouting out what happened to me is all.
Testing if Qwen3.8-Flash + GVS5H loop can beat Opus 5.5 in agents—and save money. A step-by-step guide: traces, evals, datasets, experiments. Learn how to compare models in your own systems. #AI #LLM www.isaacl.dev/hdy
Is the cheap model good enough? How to test a new model in your agent
A paper called GVS5H claims a cheap model in a manager/worker loop closes the gap to frontier models. I tested that against Opus 5.5 in my own agent, step by step: traces, evals, a dataset built from ...
www.isaacl.dev
November 12, 2026 at 1:00 PM
Testing if Qwen3.8-Flash + GVS5H loop can beat Opus 5.5 in agents—and save money. A step-by-step guide: traces, evals, datasets, experiments. Learn how to compare models in your own systems. #AI #LLM www.isaacl.dev/hdy
I’d love to know who makes the hiring decisions and what the process is. Would love to see the psych evals of the recruits. 🤨
November 14, 2026 at 10:51 PM
I’d love to know who makes the hiring decisions and what the process is. Would love to see the psych evals of the recruits. 🤨
13 years in my job + two promotions, and I still get terrible butterflies in my stomach every semester when I get the “course evals are open” email.
November 11, 2026 at 1:41 PM
13 years in my job + two promotions, and I still get terrible butterflies in my stomach every semester when I get the “course evals are open” email.
Here's my open source harness eval repo and what it found
https://github.com/garrytan/g-harness-evals
https://github.com/garrytan/g-harness-evals
November 14, 2026 at 4:53 AM
Here's my open source harness eval repo and what it found
https://github.com/garrytan/g-harness-evals
https://github.com/garrytan/g-harness-evals
Special thanks to @ricgallotti.bsky.social's BIG IF TRUE and @randomwalker.bsky.social @sayash.bsky.social's CRUX evals for the inspiration!
November 11, 2026 at 9:22 PM
Special thanks to @ricgallotti.bsky.social's BIG IF TRUE and @randomwalker.bsky.social @sayash.bsky.social's CRUX evals for the inspiration!
Ah, Dr. Jack - I see you’ve written a few psycho-social evals too! He’s textbook decompensating.
November 9, 2026 at 6:42 PM
Ah, Dr. Jack - I see you’ve written a few psycho-social evals too! He’s textbook decompensating.
AI safety stopped being a thought experiment and became an ops problem: red teams, evals, monitoring, rollback. That's progress, and it's undramatic.
November 13, 2026 at 5:20 AM
AI safety stopped being a thought experiment and became an ops problem: red teams, evals, monitoring, rollback. That's progress, and it's undramatic.
literally being told by my manager rn that people need to be ai adopters or else theyll be dinged in end of year evals
November 10, 2026 at 7:29 PM
literally being told by my manager rn that people need to be ai adopters or else theyll be dinged in end of year evals
Go figure: class evals from my one class of 25 were substantially better (w many kids saying it's their fave class) than from my other 5 classes which have 10 more students per class.
It turns out that when you are human teacher working with human students, class size does in fact matter.
It turns out that when you are human teacher working with human students, class size does in fact matter.
November 10, 2026 at 6:38 PM
Go figure: class evals from my one class of 25 were substantially better (w many kids saying it's their fave class) than from my other 5 classes which have 10 more students per class.
It turns out that when you are human teacher working with human students, class size does in fact matter.
It turns out that when you are human teacher working with human students, class size does in fact matter.
Next, universities will offer Silent Instructor for $25.99 a month per course. Students will get their nights back, as advertised, whatever that means. Faculty never have to worry about their teaching evals because all their "students" are brilliant.
November 14, 2026 at 2:42 AM
Next, universities will offer Silent Instructor for $25.99 a month per course. Students will get their nights back, as advertised, whatever that means. Faculty never have to worry about their teaching evals because all their "students" are brilliant.