Tamer Prints (@PinGeekTamer) header
Tamer Prints (@PinGeekTamer) profile picture

عاشق التكنولوجيا 🖥️🎮 وساحر الكرة والدبابيس 🏆

Alexandria
Joined June 2022
Tamer Prints retweeted
the winner of an Anthropic hackathon open sourced his whole Claude Code setup, and it turns one assistant into a small engineering department 68 subagents, 286 skills, 94 commands, MIT license a blueprint comes before any build, a failing test comes before any fix, and every change gets reviewed by a context that never saw it written the part worth copying first is that fresh-context reviewer: one extra reader per diff, with no stake in the code begin with a single plan and a single rules pack. switching on all 286 skills at once is the quickest way to make it worse the same fresh-context rule, as a verifier seat next to Fable 5.2, is below:
full harness guide for Claude Fable 5.2 + jev engineering one evening with this and Claude does the work of a small dev team no $550/month course needed:
3
11
712
Tamer Prints retweeted
Join us for ipXchange's upcoming Physical AI Week — five days of practical, engineer-led sessions exploring the technologies and challenges behind real-world deployment. 🗓️ November 30th - December 4th Register now! luma.com/vgnhai1m?utm_source…
4
16
2,806
Tamer Prints retweeted
Some archaea—an ancient group of microorganisms—have an entirely novel genetic code, according to a 2025 study in Science. Their genetic code reassigns a DNA “stop” signal to incorporate the rare amino acid pyrrolysine into their proteins. The findings expand our understanding of how alternative genetic codes evolve and hint at new molecular tools for biotechnology applications. scim.ag/49xnYx3 #ScienceMagArchives
3
8
31
9,894
Designed for accurate multi-axis motion, it combines X, Y, and rotational movement in one integrated stage to support demanding semiconductor applications. 👉 Request a Quote: kensingtonlabs.com/contact-u… #Semiconductor #PrecisionMotion #MotionControl
1
2
5
28
What if your machines could keep running after first shift? 🤖 Get practical answers about machine tending, ROI and getting started with no CapEx from Mitsubishi Electric + Formic experts on Dec. 9. Register 👉 okt.to/DktW8r
2
8
130
Tamer Prints retweeted
From Human Demonstrations to Robot Training Data With RLWRLD | Cosmos Labs teapawt.org/i/broadcasts/2kEqrOvzp…
9
7
1
32
3,623
Tamer Prints retweeted
This little robotic duck has learned to clean up! WonderDuck, a bipedal robot from Hiwonder, demonstrates how reinforcement learning can help robots perform physical tasks. In this demonstration, the duck spots an object, picks it up and tosses it into a bin. Behind this playful action lies a serious robotics challenge: combining perception, balance, movement and object handling. Reinforcement learning allows robots to practise actions through trial and error, often in simulation before moving to the real world. #WonderDuck #Hiwonder #ReinforcementLearning #EmbodiedAI #Robotics @postoff25 @devaang @Nicochan33 @Fabriziobustama @mikeflache @SpirosMargaris @rwang0 @IsabellePiel29 @thierry_pires @MadiSeydi @sayedflah @patricegorissen @amit6060 @HaroldSinnott @fogoros @faustospain @BernardMarr @sonu_monika @pierrepinna @NewsNeus @1OFFGINGER @FmFrancoise
3
7
1
12
945
Best of luck to all the nominees, and we hope everyone enjoys a fantastic evening! #Apprenticeships #EngineeringCareers #ApprenticeOfTheYear #SkillsForTheFuture [3/3]
2
5
41
Tamer Prints retweeted
November Autodesk Forma Construction Releases – Built for What’s Next autodesk.com/blogs/construct… #AutodeskExpertElite
3
1
60
Tamer Prints retweeted
so maybe i just wait 1-2 more years till grok bots good enough to run a shop floor and get into it 🤔
3
2
3
501
Tamer Prints retweeted
Elon himself reports that placing a colony on Mars will require one million metric tons of payload lifted far beyond Earth’s orbit.
SpaceX launched an estimated 2,330 metric tons of payload to orbit in 2025. China launched 328 metric tons, Russia 64, and Europe 16.
9
9
1
75
7,404
Tamer Prints retweeted
Robotics will never be the same again . just use the ide get the work done which used to take weeks of time .
First came vibe coding, now we have vibe robotics! 🦾✨ Today @GoogleDeepMind is releasing Gemini 4 Hydrogen + the new Robotics plugin in @Antigravity, turning AGY into a one-stop robotics workbench: 🔌 USB plug-in to a calibrated, moving robot arm in ~10 mins 🌐 MuJoCo right in the workspace (adjust physics & scenes via chat!) ⚡ Carbon-level 6-axis control at Hydrogen speed and cost Go from a robot in a box to real-to-sim-to-real in a single afternoon. Huge props to the team! 🔗developers.googleblog.com/ge…
1
8
163
11,055
Tamer Prints retweeted
Microsoft CEO, Satya Nadella says Microsoft will operate 200,000 million AI agents in loops that complete entire jobs. This is the same Microsoft that gave AI to those 200,000 people & then wrote on their own blog in September that it didn't change the work. "A tool licensed to over 200,000 people does not change how the work gets done." Everyone uses AI without having any actual impact on the P&L. I see it every day with our clients. In the same talk with Reid Hoffman he says he runs around 100 agents himself and it's already a lot. "The cognitive load on me managing this is so high." Now picture a company with thousands of these. Who owns them? Who's QAing them? Who decides what they're allowed to touch and who gets the call when one breaks on a Thursday? His words: "You need to give them identities, you need to give them sandboxes, then you need to set policies to govern them." That's pretty much how we run it. We start with back-office operations. Then, manual repetitive data entry, the handoffs and the processes in between the work. Emailing a customer or moving money still gets a yes from someone on their team. If people weren't using something, we took that as a sign we hadn't gotten it right yet. Once a few people saw their own work get easier it started to spread, people were helping us test and give feedback almost instantly. Some of our clients have doubled the accounts they can take on within 90 days without adding anyone to payroll. Adoption is definitely tricky because humans are creatures of habit. Most companies are going to buy agents the exact same way they bought seats.
19
31
112
40,799
The cleanest data result of CoRL 2026's workshop morning came from @SongShuran (@Stanford), and it indicts how most labs collect robot data. Part 3️⃣/4️⃣: robot in-context learning. @corl_conf 1️⃣ A demonstration is the prompt. Behavior prompting turns π(a|o) into π(a|o, demo). Language says what, a goal image shows the end state, and a demo specifies how: trajectory, timing, recovery. 2️⃣ DrawAnything is the right stress test. Language can't specify stroke order and curvature, and pure replay can't adapt to where the board is. Thousands of distinct tasks are cheap to generate. 3️⃣ Diversity beats repetition. Going from 125 to 2,000 training tasks cuts error from 35.6 to 3.4 px, with a phase transition near 500 tasks. At 500 tasks, 4× more demos per task moves error from 9.7 to 9.5. 4️⃣ Below the threshold, it breaks. With ~3 laundry tasks, demo prompts lose to language and competently execute the wrong fold. 5️⃣ The architecture is built for long prompts. BPP pools each ~1 Hz chunk of observations, proprio and actions into one embedding, cross-attends to them at linear cost, and decodes actions with a FiLM-conditioned diffusion policy. Error drops 80.7% vs goal-image prompting and 32% vs ICRT, competitive with fine-tuned π0.5 without any pretraining. 6️⃣ Scale gets there another way. Generalist's GEN-1.5 runs a new task from one 3–12 s "physical prompt": 59% average one-shot success across 10 tasks, 83% after 10 gradient steps on ~5 minutes of data. No meta-learning objective. My read: this morning showed three roads to robot ICL. Train for it (BPP), let it emerge at scale (GEN-1.5), or write it into fast weights at test time (RoboTTT). Song's threshold may explain the emergence: broad pretraining simply crosses the diversity bar, as LLM work on in-context regression found. Hot take: stop collecting "one task, dozens of demos." Collect many tasks, a few demos each. #CoRL2026 #InContextLearning #RobotLearning #EmbodiedAI
3
7
1
41
4,090
Tamer Prints retweeted
Getting started is as simple as teaching the robot where to go and letting it learn its surroundings. And as your production needs grow, Zimmer Group's fleet management software makes it easier to add more robots and expand your automation setup - youtube.com/watch?v=EXRid9Jv…
1
3
6
262
Tamer Prints retweeted
Built on work with @PNNLab, Abaco brings shared, rack-scale memory closer to production. See how it helps #AI and #HPC systems keep active data near compute, scale capacity independently and tackle the memory wall. Learn more: bit.ly/4zCVoE0
2
10
77
4,907
Vibe robotics is here! If you thought Astra / Carbon / Fable are too slow for controlling real robots, try Hydrogen. Hydrogen is much faster than Carbon and now there's an anti-gravity plugin where you can just connect your hardware and prompt it to do things in under 10 minutes - no fine-tuning / hosting needed. It has a bunch of skills relevant for control, simulation, harnesses and you can even practice in-context in sim before rolling out in the real world :) No better time to be building robotics applications.
Teaching robots new tasks just got a lot faster. 🤖 With Gemini 4 Hydrogen and the new robotics plugin for Google @Antigravity, you can connect a supported arm, practice in @MuJoCo simulation, and run tasks in the real world using plain conversation.
18
22
333
21,835
Tamer Prints retweeted
Really good post. As the compute bottleneck clears for training humanoids in sim, model accuracy is going to be one of the big areas where new gains will be found. I've been in several hallway discussions over the last year about this incongruence between the simulated and real G1 ankle (and alllll of the humanoids that copy the same exact linkage). This, and the work done in OmniRetarget/LadderMan for collision models are worth diving deep into.
Qiayuan (@qiayuanliao) and I wrote about modeling the Unitree G1’s closed-loop ankle linkage in @mujoco. We look at how the linkage changes the ankle’s motion and dynamics, and what it takes to simulate it reliably. We spent quite a bit of time on the blog post making interactive demos and shareable models, and we hope you find it useful! kzakka.com/closing-the-loop/
1
2
30
2,010
Tamer Prints retweeted
NEW RESEARCH: Predicting tuch is actually more useful than merely reading it! This project involves @Ismini_L and others Called DexTacWAM, for A Visuo-Tactile World-Action Model for Dexterous Manipulation, it is a method from UC Berkeley, UIUC, and Northwestern. It puts fingertip touch inside a video world model rather than beside it. Each fingertip's grayscale tactile map is painted as pseudo-RGB and pushed through a frozen pretrained VAE, the same encoder used for the camera streams and shared across all ten fingers. A compressor with finger-identity embeddings, hand-pose conditioning, self-attention pooling and temporal refinement folds five fingertip streams per hand into a single latent, cutting DiT tokens from 3,168 to 864. Those tactile latents are appended as extra views to a 2B-parameter pretrained video diffusion transformer, which is then trained with a modality-split flow-matching objective to denoise both the visual and the tactile future. An action expert cross-attends to the world model's latents with per-modality K/V RMS normalization. Training runs in three stages: 1. adapt the tactile encoder on a 488-episode, roughly four-hour bimanual corpus called Diverse-488 with the VAE frozen 2. finetune the world model on about 100 demonstrations per task 3. then train a fresh action expert on the same demonstrations. The platform is a 22-DoF bimanual rig with Sharpa hands and vision-based fingertip tactile sensors, head and wrist cameras. I find it super interesting that feeding the model touch does not bring much value, while making it predict touch actually enhances its performance.
2
8
1
33
1,906
Tamer Prints retweeted
Autonomous systems run on nanosecond-precision data. #InfluxDB handles the real-time telemetry that lakehouses can't, powering self-correcting AI from manufacturing floors to autonomous vehicles. ⚡ bit.ly/4h4eb1V by @evankaplan via @ForbesTechCncl
2
3
284