Daily Health nd Fitness Tips
Joined July 2023
- Tweets5.2K
- Following0
- Followers85.3K
- Likes51.6K
Alpha Prime retweeted
Been hearing some pretty interesting things about a small AI research team experimenting with post-training Qwen 3.8 27B.
Apparently their post-hoc GQA → MLA conversion, which reportedly reduces KV-cache memory requirements by up to 8.4x without additional training
Alpha Prime retweeted
A lab just open-sourced a 27B model built for one job: acting on unclear input without making things up. 🧠
Think about what an agent actually does. It gets a vague instruction, fills the gaps with a guess, and then acts on that guess. When it’s wrong and confident, you don’t get a bad answer. You get an action you never asked for.
That explains the metric they went after.
→ Hallucination rate 15.1%, vs 30.4% for its own base and 85.3% for GPT-6 Luna
→ Terminal-Bench 2.1: 76.0, best in their comparison set
→ τ³-bench: 259/278, next best 220
→ MCPMark: 48/55, next best 42
→ 29.3K median output tokens per task, vs the base model’s 65.3K
Half the hallucinations and half the token spend of the model it came from.
GPT-6 Luna still wins DeepSWE and the omniscience index, and general-knowledge scores sit level with the base.
So this was targeted work, not just a scaling run.
Alpha Prime retweeted
Someone built a drone with 3D-printed spiral fans instead of regular propellers.
The design keeps the blades enclosed, reduces turbulence, and can improve thrust while making the drone safer to handle.
Even the way this thing is built is worth a closer look.
Alpha Prime retweeted
Some of the most interesting agent results this year are coming out of post-trains of small open models.
A new RL post-train Qwen3.8-27B model trained with an open-source environments cuts the hallucination rate on AA-Omniscience from 30.4% to 15.1%.
the compressed copy of the mla version fits on my single 24 GB RTX 4090 and holds about 3.5x more context than base Qwen.
It beats GPT 6 Luna-medium on DeepSWE, placing in between Luna 6 medium and high.
scores higher on Terminal-Bench 2.1 while writing a median 23.8K output tokens per solved task, where base Qwen writes 54.9K.
Tried an unreleased version of this model, and really felt the comparative intelligent density, that it writes a lot less.
tops Muse Glimmer 30B and OpenAI’s GPT-6 Luna (high) on Terminal-Bench, τ³-bench, ClawEval and BFCL v4 (Berkeley Function Calling Leaderboard), though Luna still wins DeepSWE at 53.1%.
They also released an MLA version, which is the same model with its 16 full-attention layers converted to multi-head latent attention, which shrinks the memory each token takes from 64 KiB to 18 KiB so far more or longer sessions fit on one GPU.
Alpha Prime retweeted
You can now make up to $50,000/month editing videos without knowing how to edit.
Claude can now handle the editing for you.
Here are ways you can make money with it:
Alpha Prime retweeted
Everyone thinks stretching is the best way to prevent injury.
But this is a lie.
Here are 7 types to avoid:
1. Bouncing stretches
Alpha Prime retweeted
Four people start the same diet on the same day.
Same calories. Same 90 days.
All four lose weight. Only one ends up truly healthier.
Here's why:
Alpha Prime retweeted
The annoying thing about most AI assistants is that you still have to do half the work yourself.
You ask, it answers, then you’re back to opening apps and sorting things out.
SpeakON is taking a different approach.
You speak, and its personal AI agent can help with things like emails, reminders, and scheduling.
That’s the kind of AI we’d actually want in our daily routine.
we crushed a keyboard.
its retirement party got a little out of hand.
for years, it was there for every email, every plan, every “just one more thing.” a loyal colleague. a very loud colleague.
but your ideas move faster than your fingers.
and somewhere between knowing what you want to do and typing it all out, a quick task becomes another thing taking up your afternoon.
meet SpeakON AI Agent.
start with what you need, in your own words. less time behind a blinking cursor. more room to move things forward.
the keyboard had a good run.
now, let’s give your voice a turn.
THE LAST KEYSTROKE is here. watch the film, then tell us: what’s the first task you’d hand over?
speakon.app/x
Disclaimer: The calling feature is still in beta and will be available soon. All other Agent features are available now, with ongoing improvements.
Alpha Prime retweeted
Ever remember something you need to do right when you're in the middle of something else?
Like replying to an email while you're working on a project, and knowing you'll probably forget if you don't write it down.
That's what I find interesting about SpeakON. You can capture the thought by voice, then use its AI agent to help with things like reminders and emails.
One less thing to keep in your head.
we crushed a keyboard.
its retirement party got a little out of hand.
for years, it was there for every email, every plan, every “just one more thing.” a loyal colleague. a very loud colleague.
but your ideas move faster than your fingers.
and somewhere between knowing what you want to do and typing it all out, a quick task becomes another thing taking up your afternoon.
meet SpeakON AI Agent.
start with what you need, in your own words. less time behind a blinking cursor. more room to move things forward.
the keyboard had a good run.
now, let’s give your voice a turn.
THE LAST KEYSTROKE is here. watch the film, then tell us: what’s the first task you’d hand over?
speakon.app/x
Disclaimer: The calling feature is still in beta and will be available soon. All other Agent features are available now, with ongoing improvements.
Alpha Prime retweeted
Most executives think fat loss means more hours in the gym.
I dropped 22lbs of fat in 9 weeks.
Same training. Same schedule. Same workload.
The only lever I pulled was nutrition.
Here’s exactly what I ate:
Alpha Prime retweeted
Your back pain isn’t from age.
It’s from the 9 hours you spend in a chair...
And the zero minutes you spend undoing it.
Here are the 4 exercises I do every week to bulletproof my back:
= Thread =
Alpha Prime retweeted
"I eat clean and I still can't lose fat."
The most common sentence Apex hears from executives.
Eating clean is not the same as fat loss.
Two numbers run the show. Here's what they are and how to use them:
= Thread =
Alpha Prime retweeted
150 bitcoin. one locked phone. one forgotten pattern.
in 2013, a man reportedly lost access to 150 bitcoin stored on his android phone after forgetting his screen unlock pattern. what seemed like a minor mistake turned into a financial nightmare as the cryptocurrency’s value skyrocketed over the following decade.
despite years of attempts, he has reportedly been unable to unlock the device and recover his digital fortune.
the story has gained widespread attention online, serving as a reminder of how something as simple as a forgotten password can potentially lock someone out of millions of dollars in cryptocurrency.
Alpha Prime retweeted
A Chinese company built a robot that parks your car automatically, saves garage space and helps prevent scratches.
Drop off your car and walk away. The robot does the parking.
Your parking skills just became optional.
Alpha Prime retweeted
Can we get this kind of movement, but with our character?” That's a very different challenge from simply generating a clip.
Motion capture in Dreamina Canvas lets you use character motion and spatial depth from a reference video to help create a new scene.
And for just $1.50, Dreamina Basic gives you a full month to explore different creative directions.
Dreamina 2.0 is canvas-first, the new way AI video is made. dreamina.capcut.com/ai-tool/…
#Dreamina #DreaminaPartner