Rez Karim

Giving your AI agents superpowers at @unsora_ai. I talk AI video, ads & prompt engineering.

[Rez Karim @rezkhere]

Holy shit...
I think I created a monster.

1. Upload your product
2. Select from past successful ads in your niche/from competitors.
3. Get as good or better ads for your brand

Comment 'ads' and I will DM you the access.

Also share your product below. Imma make you some ads :)

[Rez Karim @rezkhere]

@ilan_kadar @pluraiAI the replay from the exact step thing alone would’ve saved me so much time.

[Rez Karim @rezkhere]

writing eval scenarios from scratch is the most boring part of shipping an agent.

Plurai reads your code and generates the personas, flows and evals for you.

then those personas push back on your agent like real users do.

wild

[Ilan Kadar @ilan_kadar]

Big day for us at @pluraiAI ! Plurai Simulation is live and free for every AI engineer.

AI agents look great in demos. Until real users break them.

So we built a world where your agent can fail, learn, and get fixed before your users ever see it.

→ Digital twin: Your real agent, its tools, and environment recreated from your code

→ Simulated users & tasks: Realistic users, tasks, and workflows that push your agent into the edge cases that matter

→ Auto evals: Check what your agent actually did, not just what it said

→ Verified fixes: Every failure comes with the run that exposed it and a fix you can verify

→ Replays: Turn any production failure into a simulation and replay it against every new version of your agent

All inside the coding agent you already use: @claudeai code, @OpenAICodexCli , or @GitHubCopilot .

Start simulating for free: https://t.co/1PNPQdW29S

[Rez Karim @rezkhere]

@treg_ai The fact that the whole pipeline can now be automated.

[Rez Karim @rezkhere]

my AI UGC workflow rn:

- have claude pull the trending tiktoks in my niche with treg

-read every transcript, break down what the hooks and characters have in common.

- then build an original character and turn her into UGC clips with Seedance 2.5.

- find the winners and double down

Its that simple.

[Treg @treg_ai]

OpenSource our AI-UGC workflow

For as low as $0.2/s instead of $50 per video

Try it here: https://t.co/DBdmmcWHHy

Reverse-engineer trending videos to scalable UGCs that brings $$$ millions of impressions

Repo link below 👇

[Rez Karim @rezkhere]

The question this post answers for agent builders:

where does the agent go to try things safely?

Into a model of the world. Open, downloadable, post-trainable.

[Eric Xing @ericxing]

Excited to share our latest edition of world model PAN 2 built on a Generative Latent Prediction (#GLP) architecture. A world model needs to operate in an "Aristotelian world" instead of a "Platonic world" where physical intelligence translates to 6 dimensions of capability: obeying the laws of physics, coherence over history, persistence beyond sight, conformity to constraints, invariance to viewpoints, adherence to objective dynamism; and furthermore provides 3 advanced functions: acting as a world simulator that is steerable with actions and plans (long sequence of sections), thereby enabling stimulative reasoning or thought-experiments, and therefore supporting agent to solve complex tasks, and doing so in a fully auditable fashion (hence Aristotelian). A full paper will be posted soon to explain in details how we achieve this.

Congratulation to the @IFM_AI PAN team.

[Rez Karim @rezkhere]

@ericxing is predicting the next world meant to cover sound too, or just what you can see?

[Rez Karim @rezkhere]

too many flashy demos these days. few can pull it off live.

when fish audio said it’s fast, wasn’t expecting it to be this FAST lol

[Rissa Cao @rissa_cao]

Introducing @FishAudio Live 3: a full-duplex voice model that listens, says "mm-hmm," and knows when to shut up.

Enterprise? Book a demo for your free custom agent: https://t.co/2kyRhcFMxM

Reply with your use case - we'll build it free for the first 1,000 people.

[Rez Karim @rezkhere]

Transcript edits take formatting rules written in plain English, like 'write every date as YYYY-MM-DD'.

I've written that exact rule as a post-processing step at three different jobs

[ElevenLabs @ElevenLabs]

Introducing Scribe v3 and Scribe v3 Live: our most accurate Speech to Text models yet.

Built for real-world audio, Scribe excels in situations with extreme background noise, knows pauses from the end of sentences, and keeps track of who said what in a busy meeting.

[Rez Karim @rezkhere]

been waiting for someone to do ai agents in imessage properly.

most of them can chat and that’s about it.

this one actually does stuff.

- sum up a long reddit thread
- fix a old photo before i post it
- make a quick video from one text

all from a text thread.

pally x monid is a good combo ngl

[Shengkun Ye @shengkunye]

Introducing the most powerful iMessage agent.

Your iMessage now has 3,000+ tools to:

> Read YouTube, Instagram, TikTok, Reddit, and LinkedIn
> Edit your photos
> Generate music and videos
> Check stock and crypto prices
> Pull job openings

@Pally is now powered by @MonidHQ.

[Rez Karim @rezkhere]

been testing voice agents for a while and the thing that always messes it up is interrupting them.

they either ignore you or completely lose the thread. duplex is the fix, it’s listening the whole time it’s speaking.

fish’s live3 took a bakery call like a real front desk.

and the wildest part is it was built in under 15 seconds

[Rez Karim @rezkhere]

been testing voice agents for a while and the thing that always messes it up is interrupting them.

they either ignore you or completely lose the thread. duplex is the fix, it’s listening the whole time it’s speaking.

fish’s live3 took a bakery call like a real front desk.

and the wildest part is it was built in under 15 seconds

[Rissa Cao @rissa_cao]

Introducing @FishAudio Live 3: a full-duplex voice model that listens, says "mm-hmm," and knows when to shut up.

Enterprise? Book a demo for your free custom agent: https://t.co/2kyRhcFMxM

Reply with your use case - we'll build it free for the first 1,000 people.

[Rez Karim @rezkhere]

@FishAudio @rissa_cao Lessgo 💯

[Rez Karim @rezkhere]

@meetbackdrop technical founders who hate sales calls will love this ngl

[Rez Karim @rezkhere]

your prospect just asked something you can't answer

you have 3 seconds

and backdrop already has it on your screen

nothing joins the call. you get just the answer

feels like chatgpt but for sales calls

wild

[Backdrop @meetbackdrop]

Selling is hard. Live calls are brutal. Forget one question and the deal might be gone.

Your buyers are impatient. Someone pitches them a version of your product every single day.

Until now, AI couldn't help you in the moment. Too slow. Too distracting.

Not anymore.

Introducing Backdrop: the first live sales assistant that doesn't suck.

Rookie or veteran, it doesn't matter. Backdrop has already helped dozens of teams close 6-figure deals.

[Rez Karim @rezkhere]

- small open model
- one specific job

small teams often have the founder making a campaign in one sitting and hoping it works.

a tool like this is built for exactly that person.

[Rez Karim @rezkhere]

@ginacostag_ This is huge milestone for the space.

[Rez Karim @rezkhere]

- small open model
- one specific job

small teams often have the founder making a campaign in one sitting and hoping it works.

a tool like this is built for exactly that person.

[Ted @notmiramurati]

Meet Yvligo, our open-weights model for running X ads.
(3x better than Grok 4.7)

At https://t.co/pAn5O703dD, we built an X simulation and post-trained DeepSeek V4.1 Flash, outperforming frontier models

Live results:
Yvligo: 383 relevant website visits for $200
Grok: 123 visits

[Rez Karim @rezkhere]

@notmiramurati congrats man. open weights is a big deal for this ngl

[Rez Karim @rezkhere]

Nearly 5,000 hours of touch recordings, from gloves on people's hands and from robot hands.

Somebody sat through a lot of grasping to get that.

[Robbyant @robbyant_brain]

What changes when robots can feel—not just see? 🤖✋
Meet LingBot-Touch, our open-source tactile foundation model towards dexterous manipulation, pretrained on nearly 5,000 hours of real tactile data.
Through unified heterogeneous tactile modeling, LingBot-Touch supports transfer across sensors, dexterous hands, tasks, and action models.
Watch it in action. 👇 #LingBot #EmbodiedAI #Robotics