osama
@_o_s_a_m_a
Followers
4K
Following
37K
Media
20
Statuses
291
ai/robotics nerd. cpto @durableai. prev built things at burrata, segment, streamlabs, gatech š
Vancouver
Joined June 2008
wrote an ampcode prompt so long that it consumed 10% claude context window on cmd+enter.
0
0
0
product mgmt today is where project/program mgmt was ~10ā15y ago. as pm/em absorbed project mgmt, design/eng/data/customer success will absorb āproduct sense + execution.ā . ai eats coordination ā product sense moves to the edges (design, eng, data, customer success). slowly,.
We asked @clairevo (Founder, ChatPRD) about the future of product management and how AI tools could reshape the role entirely. Vo predicts a significant shift, suggesting traditional PM roles won't last in their current form:. āProduct managers are either going to develop the
1
0
1
mech suit for codegen models. curious to try.
Weāve been working with OpenAI over the last few weeks on GPT-5 in Amp. Weāre still evaluating where and how to use GPT-5 (after all, it took 6 months to figure out 3.5 Sonnetās true capabilities). Join us and try GPT-5 in Amp now:
1
2
13
RT @nikunj: Mini rant on how weāve swung the pendulum a bit too far from product-maxxing to views-maxxing. For years, companies have beenā¦.
0
36
0
good callout.
World modeling for robotics is incredibly hard because (1) control of humanoid robots & 5-finger hands is wayyy harder than ā¬ļøā¬
ļøā¬ļøā”ļø in games (Genie 3); and (2) object interaction is much more diverse than FSD, which needs to *avoid* coming into contact. Our GR00T Dreams work was.
0
0
0
genie3 accelerates robotics/physical-ai timelines by 1-2years if not more. rl env going to get lit. humanoids arriving sooner than 2030. lfg!. most underrated release today: . note to self: buy more google stock.
deepmind.google
Today we are announcing Genie 3, a general purpose world model that can generate an unprecedented diversity of interactive environments. Given a text prompt, Genie 3 can generate dynamic worlds...
0
0
4
oai cooked. first american oss model.to be steps ahead and not just incremental gains.
Here's the concluding section of my write-up of the new OpenAI open weights models - they're really impressive, and I think they hold their own (or even beat) the models we've seen come out of the various Chinese AI labs over the past few months
0
0
1
RT @willdepue: less than one year between o1 announced (september 2024) and we have an o3-level model open sourced (august 2025) thatās runā¦.
0
26
0
apple doesn't need to build a foundation model. all they need to do is pack an m4max+battery into a portable device and embed the gpt-oss into coreml dx such that it can be finetuned.
0
0
2
RT @FFmpeg: It's important to remember that "technology influencers" can be bad role models. Many encourage people to be "fans of technologā¦.
0
2K
0
good read for everyone building cyber physical systems. blake writes, i repost. no. 1 fan.
0
0
2
folding shirts is to world foundation models as fart jokes were to large language models? . this talk is a good survey. it feels like we're within 2-3 year range of chatgpt moment for physical ai. 2030 will be lit for robotics.
1
1
2
many members of our eng team have started using @AmpCode within and outside cursor. it is objectively better than all other out there today. we've also got it running inside pre-commit hook (with Playwright mcp etc) and doing pr review assists. with oracles and subagents, it.
Many coding agents have a "modes" feature that (kinda?) made sense in a RAG or early agentic era. Now they're adding subagents like Amp (search subagent, general subagents, the oracle) and Claude Code (which just shipped a really cool custom subagents feature). Modes are.
3
6
45
RT @SullyOmarr: by far the best sdk for ai (typescript) and its not even close. its so good that if you're not using this you're actively hā¦.
0
9
0
solid release. no surprise that the neatest abstraction for ai engineering is coming from application devs and is in ts. we use v5 is prod. come join the most fun ai playground / candy land for engineers at @DurableAI.
AI SDK 5. Introducing type-safe chat, agentic loop controls, data parts, speech generation and transcription, Zod 4 support, global provider, and raw request access.
0
0
9
1. know thyself .2. know thy game.
It feels like there are two modalities of competition in Silicon Valley today. Both are valid and interesting and can lead to big outcomes. Modality 1: Work on something legible. Out execute the fuck out of everyone. --> This means that you are working on something that is.
0
0
0
RT @mitchellh: Be wary of people who spew AI hype and screenshot their army of agents but don't seem to be shipping anything but demos. I'mā¦.
0
143
0