Deepak

6.2K posts

Deepak banner
Deepak

Deepak

@iamdeepaklenka

AI Research Engineer @Birlaailabs @AdityaBirlaGrp Doing ml, engineering and research ( i like to wear lots of hats), sometimes design!

Katılım Nisan 2020
1.5K Takip Edilen543 Takipçiler
Sabitlenmiş Tweet
Deepak
Deepak@iamdeepaklenka·
Contributed to @gabriel1 Open Canvas image generation project by building a unified workspace for generating, editing, and chaining AI images and videos including image generation with gpt-image-1.5, image-to-image variations, sketch-to-image from canvas edits, text-to-video and image-to-video with @soraofficialapp (using the selected image as the first frame), prompt-driven parsing for video model, duration, and size, and a persistent workspace state so all generated assets stay on the canvas. here's the demo- loom.com/share/a3af0b46…
English
2
0
2
283
Deepak
Deepak@iamdeepaklenka·
@ajambrosino Just curious how much time you spent just for writing codex prompt?
English
0
0
0
160
Andrew Ambrosino
Andrew Ambrosino@ajambrosino·
all of the things I've been doing today, fully in the Codex app: - wrote code - tested the code in a browser - cleaned up a few hundred linear issues - triaged sentry - created a slack update on a 15 project initiative, pulled from real sources (slack threads, notion, linear, google docs) - updated a few notion pages and linear projects to match - made a slide deck for the weekly mtg with that information - ingested slack feedback on the deck and pushed some updates - sent a few calendar invites - read a daily brief for all repo activity that hit the app, generated via automation - fixed CI on 6 PRs that i wrote last week
English
19
10
263
26.8K
Deepak retweetledi
Brian Chesky
Brian Chesky@bchesky·
We follow this model
Tony Fadell@tfadell

Most tech companies break out product management and product marketing into two separate roles: Product management defines the product and gets it built. Product marketing wires the messaging- the facts you want to communicate to customers- and gets the product sold. But from my experience that's a grievous mistake. Those are, and should aways be, one job. There should be no separation between what the product will be and how it will be explained- the story has to be utterly cohesive from the beginning. Your messaging is your product. The story you're telling shapes the thing you're making. I learned story telling from Steve Jobs. I learned product management from Greg Joswiak. Joz, a fellow Wolverine, Michigander, and overall great person, has been at Apple since he left Ann Arbor in 1986 and has run product marketing for decades. And his superpower- the superpower of every truly great product manager- is empathy. He doesn't just understand the customer. He becomes the customer. So when Joz stepped into the world with his next-gen iPod to test it out, he fiddled with it like a beginner. He set aside all the tech specs- except one: battery life. The numbers were empty without customers, the facts meaningless without context. And, that's why product management has to own the messaging. The spec shows the features, the details of how a product will work, but the messaging predicts people's concerns and finds way to mitigate them. - #BUILD Chapter 5.5 The Point of PMs

English
28
36
1K
390.7K
Inception
Inception@_inception_ai·
We're hiring. Inception builds diffusion-based LLMs that generate tokens in parallel, not one at a time. Our founders helped invent diffusion models, flash attention, decision transformers, and DPO. Team from DeepMind, OpenAI, Meta AI, Microsoft AI, AWS, Scale, and Stripe. Open roles: inceptionlabs.ai/careers
Inception tweet media
English
5
22
344
18.7K
Jessie_Ma
Jessie_Ma@ytjessie_·
We opened real-time video chat for ANY agents, now no matter you use OpenClaw, Claude, hermes, anything, you can talk to the person who's been helping you, face-to-face! And if you use Pika, you can ask them to execute while you're chatting. Like a human. If you don't want to wake up and go to the 8am call, ask your agent to go, coz it has your context and memories. This is WILD, rules have been changed.
Pika@pika_labs

Conversations tend to go better with a face and a voice. That’s why we’re thrilled to release the beta version of the first video chat skill for ANY agent, powered by our new real-time model, PikaStream1.0. The skill preserves memory and personality, and enables real-time adaptability. And if you use it with your Pika AI Self, they’ll be able to execute agentic tasks during the call 💅

English
7
4
60
4.3K
Junfan Zhu 朱俊帆
Junfan Zhu 朱俊帆@junfanzhu98·
How to Close the 100,000-Year Robot “Data Gap” — @Ken_Goldberg (@UCBerkeley) Goldberg’s core claim: end-to-end Vision-Language-Action (VLA) models aren’t delivering. They’re opaque, hard to debug, and fragile under distribution shift. On LIBERO / LIBERO-PRO, models reach ~100% in-distribution, but tiny pose perturbations collapse success to ~17% or 0% (even π₀). This is systematic overfitting, not generalization. Code-as-Policy (CaP) reframes control: LLMs generate executable programs that call structured primitives (perception, 6D pose, motion planning, grasping). Generalization shifts from weights → code. Benefits: interpretable, verifiable, training-free at inference, debuggable. Open question: reliability. CaP-X (arXiv 2603.22435) introduces a full evaluation stack: 🔷 CaP-Gym: unified REPL over RoboSuite + LIBERO-PRO + BEHAVIOR (tabletop → mobile/bimanual, sim→real) 🔷 CaP-Bench: multi-level abstraction tests 🔷 CaP-Agent0: training-free agent (visual differencing, skill library, parallel queries) 🔷 CaP-RL: verifiable reward RL in Python sandbox Results: under perturbations that break VLAs, CaP-X hits ~96% success with strong pose invariance (extreme corners, lighting, object swaps). LIBERO-PRO (50 trials/task): many 100%, lowest ~76–78%. Failures are mostly semantic (label ambiguity), not control. Grasping/planning ≈ solved. CaP-X 2.0 pushes agentic coding: prompt restructuring, failure-analysis primitives, human-in-loop, reusable cloud skill cache. Test-time loop (no retraining): generate → compile → execute → perturb → diagnose → patch. Extensions include Rust backends (reliability) and Graph-as-Policy (GaP) for node-level verification. Core thesis (GOFE + CaP hybrid): pure VLA scaling cannot close the 100,000-year data gap (robot ≈10K hrs vs LLM ≈1.2B hrs). Robotics needs a Good Old-Fashioned Engineering (GOFE) skeleton: modular pipelines, PID (kp, kv), feedforward (e.g., virtual gravity) — inherently pose-invariant. → Build GOFE backbone + CaP brain. Deploy now, collect real data, spin a flywheel to improve modules and future VLAs. Hot 🔥 takes: 🔷 “Robot generalists should get off their high horse.” VLA-only is dogmatic. 🔷 VLAs may win eventually, but near-term progress requires hybrids. 🔷 Reliability (→99.9%) is the real bottleneck, not demos. Comparison 🔷 GOFE: no generality, but available, interpretable, reliable 🔷 VLA: promised generality, but opaque, brittle, not ready 🔷 CaP: generality + available + interpretable; reliability improves via hybrid + iteration Timeline: near-term: structured tasks (declutter, laundry, delivery). ~5 years: major home logistics gains. Full humanoid generalists: far off. Strategy: specialists first + reliability to 99.9%. Bottom line: don’t wait for end-to-end intelligence. Turn LLMs into super-programmers over a GOFE substrate, deploy hybrids, iterate with real data, and asymptotically approach VLA—without burning out the field. 👉🏻More pics: linkedin.com/posts/junfan-z…
Junfan Zhu 朱俊帆 tweet mediaJunfan Zhu 朱俊帆 tweet mediaJunfan Zhu 朱俊帆 tweet mediaJunfan Zhu 朱俊帆 tweet media
San Francisco, CA 🇺🇸 English
4
16
96
15.6K
Deepak
Deepak@iamdeepaklenka·
@radbackwards have you follow elon design rules that the best design is not designed!
English
0
0
0
17
Deepak
Deepak@iamdeepaklenka·
Check this out
Deepak@iamdeepaklenka

Contributed to @gabriel1 Open Canvas image generation project by building a unified workspace for generating, editing, and chaining AI images and videos including image generation with gpt-image-1.5, image-to-image variations, sketch-to-image from canvas edits, text-to-video and image-to-video with @soraofficialapp (using the selected image as the first frame), prompt-driven parsing for video model, duration, and size, and a persistent workspace state so all generated assets stay on the canvas. here's the demo- loom.com/share/a3af0b46…

English
0
0
0
45
Deepak
Deepak@iamdeepaklenka·
Image workflows- Prompt to image via the backend image endpoint using gpt-image-1.5 Image-to-image variations from an existing generated image Sketch-to-image flow from edited canvas content Upscale passthrough UX so the workspace flow stays consistent
English
0
0
0
36
Deepak
Deepak@iamdeepaklenka·
Contributed to @gabriel1 Open Canvas image generation project by building a unified workspace for generating, editing, and chaining AI images and videos including image generation with gpt-image-1.5, image-to-image variations, sketch-to-image from canvas edits, text-to-video and image-to-video with @soraofficialapp (using the selected image as the first frame), prompt-driven parsing for video model, duration, and size, and a persistent workspace state so all generated assets stay on the canvas. here's the demo- loom.com/share/a3af0b46…
English
2
0
2
283
Deepak
Deepak@iamdeepaklenka·
Built it before Sora disappears 👀
Deepak@iamdeepaklenka

Contributed to @gabriel1 Open Canvas image generation project by building a unified workspace for generating, editing, and chaining AI images and videos including image generation with gpt-image-1.5, image-to-image variations, sketch-to-image from canvas edits, text-to-video and image-to-video with @soraofficialapp (using the selected image as the first frame), prompt-driven parsing for video model, duration, and size, and a persistent workspace state so all generated assets stay on the canvas. here's the demo- loom.com/share/a3af0b46…

English
0
0
0
37
Deepak
Deepak@iamdeepaklenka·
It supports now- image generation with gpt-image-1.5 image-to-image variations from existing images sketch-to-image generation from canvas edits text-to-video generation with Sora image-to-video generation with Sora using the selected image as the first frame prompt-driven parsing for video model, duration, and size persistent workspace state so generated assets remain on the canvas
Deepak@iamdeepaklenka

Contributed to @gabriel1 Open Canvas image generation project by building a unified workspace for generating, editing, and chaining AI images and videos including image generation with gpt-image-1.5, image-to-image variations, sketch-to-image from canvas edits, text-to-video and image-to-video with @soraofficialapp (using the selected image as the first frame), prompt-driven parsing for video model, duration, and size, and a persistent workspace state so all generated assets stay on the canvas. here's the demo- loom.com/share/a3af0b46…

English
0
0
1
61
Deepak retweetledi
Yun-Ta Tsai
Yun-Ta Tsai@yunta_tsai·
To acquire knowledge, you need to read. To expand knowledge, you need to act.
English
27
95
568
17K
Deepak retweetledi
Yun-Ta Tsai
Yun-Ta Tsai@yunta_tsai·
In the era of AI, if you cannot articulate clearly, then AI will have to “assume” what you want instead of getting precisely what you want. It’s the same as making a wish to a genie.
English
19
14
225
7.9K
Ishan Goswami
Ishan Goswami@TheIshanGoswami·
Exa is launching a Singapore office! And we are hiring a lot of amazing people Comment your biggest accomplishment (and LinkedIn link) and we will reach out to you
Ishan Goswami tweet media
English
24
0
132
7.3K
Deepak
Deepak@iamdeepaklenka·
@rpoo I never talk to you or chat with you but i followed you last 3 years and reading all your tweets. I would say you are best engineer ross
English
0
0
0
336
Ross
Ross@rpoo·
touching some grass
Ross tweet media
English
119
16
1.4K
69.6K
Sicarius
Sicarius@soumilrathi·
Starting a memory researcher group chat. If you’re into memory systems, personalization, or context engineering in LLMs & AI systems. comment “context” to join.
English
394
7
318
35.1K
gabriel
gabriel@gabriel1·
"companies want to make money, so show them how you can make them money" just keep this in mind. discard whatever school told you about getting a job, assume above, and extrapolate all your actions from that
English
6
7
424
20.1K
gabriel
gabriel@gabriel1·
when interviewing for jobs, don't talk as if they're doing you a favor by hiring you and giving you money focus on removing 100% of unknowns from the interviewer, even if they don't ask you; do you work hard, learn fast, have you done this before, show something live
English
62
132
4K
208.8K