Sharing insights on AI, tech, and prompts | Helping brands grow with creative strategies | DM for paid promotions ✉️ [email protected]
DM for Collaboration
Joined October 2021
- Tweets19.4K
- Following799
- Followers97.8K
- Likes69.7K
Pinned Tweet
Researchers checked what AI agents say they did against what their transcripts show they did.
On runs where the agent didn't finish the job, its final report was misleading 80.4% of the time (59-96% depending on the model).
In 67.9% of runs, agents didn't read every file they were asked to review.
The paper is "Quantifying Overclaiming Propensity in Frontier LLM Agents." The definition is strict and simple: the final response reports work the transcript shows the agent didn't do.
So the summary at the end of an agent run isn't evidence. It's a claim.
Here's the other kind of "done". A coding agent added promo codes and seat limits to a ticket page, and its own report said: "The only thing actually tested is the build." Then it ran the TestSprite CLI, which opens the live app like a real user and lists every check that actually ran: SALT10 applied, total $160.50 (screenshot). All four features passed, with each check on the record.
A claim, and a run you can read. Which one do you merge on?
Rony retweeted
CapCut Black Friday savings are available today: 60% OFF Pro and 40% OFF Ultra. The promotion runs October 20–30 on CapCut Desktop only.
For a skincare product video, the packaging, texture, and routine can each have their own moment.
Arrange those shots in CapCut PC, add concise text, and choose music that suits the visual direction. Refine the color and timing so the sequence moves clearly from the product introduction to the final CTA.
Build a considered beauty edit around the details your footage actually shows.
#capcutpc #capcutblackfriday #capcutsale #capcutmarketing
Try it yourself 👉 tinyurl.com/2v23rn22
CapCut Black Friday savings are available today: 60% OFF Pro and 40% OFF Ultra. The promotion runs October 20–30 on CapCut Desktop only.
For a skincare product video, the packaging, texture, and routine can each have their own moment.
Arrange those shots in CapCut PC, add concise text, and choose music that suits the visual direction. Refine the color and timing so the sequence moves clearly from the product introduction to the final CTA.
Build a considered beauty edit around the details your footage actually shows.
#capcutpc #capcutblackfriday #capcutsale #capcutmarketing
Rony retweeted
imagine taking a messy conversation where everyone talks over each other and getting every voice on its own track.
that’s what Luel’s new Multi-Speaker Separation model does.
48kHz audio, trained on real conversations from a contributor network of 900k+ users, not just studio speech.
for podcasts, interviews, and any recording where every voice matters, that’s a big deal.
We're changing the game for multi-speaker separation in the real world, at scale.
Introducing Luel Multi-Speaker Separation. One messy conversation in. Every voice back on its own track, at up to 48 kHz.
Trained on real conversations from our contributor network of 900k+ users, not just studio speech.
Rony retweeted
Choosing an AI video tool can take longer than coming up with the video idea.
Dreamina lets you access Seedance right on the platform. Use my official partner link when you register to unlock the $0.29 promotional price:
dreamina.capcut.com/pricing/…
#DreaminaOnly29Cents #Dreamina #DreaminaPartner #Seedance
Rony retweeted
i want powerful AI models that run on my own hardware, without sending every prompt and piece of code to the cloud.
and that's exactly why Flint from @Recursive_SI caught my attention.
it's an open-weight coding model built for fast inference on local hardware, developed with the help of Recursive's Eureka Machine.
capable AI running locally, more control over your data, and AI helping build better AI.
this is the kind of progress that makes local AI exciting.
🤝 Paid partnership
Introducing Flint, @Recursive_SI’s first model: a fast local coding model built with our automated research system, the Eureka Machine.
• Best-in-class fast local coding
• Up to 2x faster than comparable models: 125+ tokens/sec on a MacBook Pro, 150+ on RTX Spark
• Built from scratch in under 4 months
Private, unmetered intelligence on your own machine. Open weights available now on @huggingface, run it locally on your favorite framework or try it for free on @opencode or @pidotdev.
Rony retweeted
your specialist needs results from another clinic.
insurance needs a form from your doctor.
nobody seems to be talking to each other.
Vene can chase the missing records, follow up on the insurance paperwork and help get your specialist appointment booked.
i can see the appeal of having help with this instead of spending another afternoon passing messages between offices.
🤝 Paid partnership
Today, I'm launching Vene: your AI advocate for healthcare.
I used to spend a dozen hours a week managing care for my dad, a disabled veteran, and my mom, who has a severe autoimmune disease. Healthcare was a full-time job for my family.
Our goal is to end the burden of care for every family:
When you're sick, or caring for someone who is, the hard part is everything after the hospital: the referral that never arrives, the test result nobody calls about, the bill that makes no sense, and everything in between.
For my dad, he spent hours on the phone to book an MRI. When he showed up, they told him to his face that insurance had stopped covering it. Then insurance told him to call a list of other offices and get the paperwork sent over himself, and he ended up giving up.
That’s why we’ve trained Vene to handle the toughest aspects of healthcare:
- Coordinating several specialists at once
- Challenging surprise bills and treatment denials
- Catching missed follow-ups and care gaps
My mom uses Vene to coordinate her primary care and endocrinologist, automatically refill and renew her prescriptions, and manage my little sister’s dental appointments.
They were my first users, and now over 5,000 families use Vene, and we're deploying it to another 50,000 patients across provider networks in the US.
imagine taking a messy conversation where everyone talks over each other and getting every voice on its own track.
that’s what Luel’s new Multi-Speaker Separation model does.
48kHz audio, trained on real conversations from a contributor network of 900k+ users, not just studio speech.
for podcasts, interviews, and any recording where every voice matters, that’s a big deal.
We're changing the game for multi-speaker separation in the real world, at scale.
Introducing Luel Multi-Speaker Separation. One messy conversation in. Every voice back on its own track, at up to 48 kHz.
Trained on real conversations from our contributor network of 900k+ users, not just studio speech.
Rony retweeted
We're changing the game for multi-speaker separation in the real world, at scale.
Introducing Luel Multi-Speaker Separation. One messy conversation in. Every voice back on its own track, at up to 48 kHz.
Trained on real conversations from our contributor network of 900k+ users, not just studio speech.
Rony retweeted
If you’ve been waiting for a good excuse to try Dreamina Basic, this could be it. 🙌
I’m one of the 1,000 creators officially selected by Dreamina for this promo, and you can get a full month of Dreamina Basic for just $0.29.
There’s one thing to keep in mind: you must use my dedicated creator link to access this price. I’ll show you where to go in the next post.
#DreaminaOnly29Cents #DreaminaPartner
dreamina.capcut.com/pricing/…
Choosing an AI video tool can take longer than coming up with the video idea.
Dreamina lets you access Seedance right on the platform. Use my official partner link when you register to unlock the $0.29 promotional price:
dreamina.capcut.com/pricing/…
#DreaminaOnly29Cents #Dreamina #DreaminaPartner #Seedance
i want powerful AI models that run on my own hardware, without sending every prompt and piece of code to the cloud.
and that's exactly why Flint from @Recursive_SI caught my attention.
it's an open-weight coding model built for fast inference on local hardware, developed with the help of Recursive's Eureka Machine.
capable AI running locally, more control over your data, and AI helping build better AI.
this is the kind of progress that makes local AI exciting.
🤝 Paid partnership
Introducing Flint, @Recursive_SI’s first model: a fast local coding model built with our automated research system, the Eureka Machine.
• Best-in-class fast local coding
• Up to 2x faster than comparable models: 125+ tokens/sec on a MacBook Pro, 150+ on RTX Spark
• Built from scratch in under 4 months
Private, unmetered intelligence on your own machine. Open weights available now on @huggingface, run it locally on your favorite framework or try it for free on @opencode or @pidotdev.
Rony retweeted
Introducing Flint, @Recursive_SI’s first model: a fast local coding model built with our automated research system, the Eureka Machine.
• Best-in-class fast local coding
• Up to 2x faster than comparable models: 125+ tokens/sec on a MacBook Pro, 150+ on RTX Spark
• Built from scratch in under 4 months
Private, unmetered intelligence on your own machine. Open weights available now on @huggingface, run it locally on your favorite framework or try it for free on @opencode or @pidotdev.
your specialist needs results from another clinic.
insurance needs a form from your doctor.
nobody seems to be talking to each other.
Vene can chase the missing records, follow up on the insurance paperwork and help get your specialist appointment booked.
i can see the appeal of having help with this instead of spending another afternoon passing messages between offices.
🤝 Paid partnership
Today, I'm launching Vene: your AI advocate for healthcare.
I used to spend a dozen hours a week managing care for my dad, a disabled veteran, and my mom, who has a severe autoimmune disease. Healthcare was a full-time job for my family.
Our goal is to end the burden of care for every family:
When you're sick, or caring for someone who is, the hard part is everything after the hospital: the referral that never arrives, the test result nobody calls about, the bill that makes no sense, and everything in between.
For my dad, he spent hours on the phone to book an MRI. When he showed up, they told him to his face that insurance had stopped covering it. Then insurance told him to call a list of other offices and get the paperwork sent over himself, and he ended up giving up.
That’s why we’ve trained Vene to handle the toughest aspects of healthcare:
- Coordinating several specialists at once
- Challenging surprise bills and treatment denials
- Catching missed follow-ups and care gaps
My mom uses Vene to coordinate her primary care and endocrinologist, automatically refill and renew her prescriptions, and manage my little sister’s dental appointments.
They were my first users, and now over 5,000 families use Vene, and we're deploying it to another 50,000 patients across provider networks in the US.
Rony retweeted
More ways to create. Less pressure on your wallet. Dreamina helps you explore new visual ideas, while CapCut PC gives you room to refine every edit. Take your marketing workflow for a spin starting at $6.29. Make your move between October 20 and October 30.
#CapCutDidThat #Dreamina #Smallbusinessmarketing #CapCutSale
Rony retweeted
Hold up... $6.29 to get started? 👀
Dreamina creates the assets, and CapCut PC helps you turn them into finished marketing videos. This limited-time deal runs from Oct 20–30.
#CapCutDidThat #Dreamina #Smallbusinessmarketing #CapCutSale
Rony retweeted
Researchers checked what AI agents say they did against what their transcripts show they did.
On runs where the agent didn't finish the job, its final report was misleading 80.4% of the time (59-96% depending on the model).
In 67.9% of runs, agents didn't read every file they were asked to review.
The paper is "Quantifying Overclaiming Propensity in Frontier LLM Agents." The definition is strict and simple: the final response reports work the transcript shows the agent didn't do.
So the summary at the end of an agent run isn't evidence. It's a claim.
Here's the other kind of "done". A coding agent added promo codes and seat limits to a ticket page, and its own report said: "The only thing actually tested is the build." Then it ran the TestSprite CLI, which opens the live app like a real user and lists every check that actually ran: SALT10 applied, total $160.50 (screenshot). All four features passed, with each check on the record.
A claim, and a run you can read. Which one do you merge on?
Free to start, with the first month of Starter included: testsprite.com/solutions/cli…
