Harrison Kinsley retweeted
With the explosion of Microduck and Reachy Mini, we'll need more people to join the @pollenrobotics @huggingface teams in all functions.
If you're excited about democratizing robotics AI with open-source, my DMs are open!
If you're excited about democratizing robotics AI with open-source, my DMs are open!
71
56
848
48,552
more work on the @asimovinc fall recovery functionality.
now i need to add some sort of object throwing to the UI, the forces are invisible at the moment to push him over.
also wondering about rewarding for not just totally face-smashing. Might be helpful on the real thing.
now i need to add some sort of object throwing to the UI, the forces are invisible at the moment to push him over.
also wondering about rewarding for not just totally face-smashing. Might be helpful on the real thing.
8
3
58
4,441
Can anyone else corroborate this?
Post 2 exposing Kimi AI
That's the creators campaign docs I got from Kimi AI after their Kimi K3.1 pro release and thats the narrative they want creators to push, of course I'm out and I'm never gonna say something I don't test myself and give my honest opinion about and I don't this China vs US political situation.
17
1
118
26,039
A GPU temp reading is a probe. It's a reference point built around various assumptions. There are many important things on that card you can't even probe, because the assumption is there's proper airflow for how this card was designed. Even just how that fan is going to pick up air and move it across all the components is drastically altered by having it sit right on top of another card.
The workstation card was simply not built for this thermal/air profile.
There's a long line of people who have stacked rtx pro 6k workstation cards only to reverse course in short order, but seeing other people do this, and arguing in favor for it, convinces more people to try this for a time and cook their GPUs and various things on their motherboard.
Do not do this to $80,000 in GPUs or your surrounding $15,000 computer. Get a couple mcio risers. Lots of cases give you cool ways to still mount more cards like this in the case and keep it clean.
Like why gimp the cards && bake them? there's no reason to do this. If you MUST stack and MUST make it pretty, use water or MaxQ (even these have trouble stacking).
I say this as someone who has done lots of even sillier and dumber things, and will continue to, but I'll do my best to help others not make the same mistakes.
The workstation card was simply not built for this thermal/air profile.
There's a long line of people who have stacked rtx pro 6k workstation cards only to reverse course in short order, but seeing other people do this, and arguing in favor for it, convinces more people to try this for a time and cook their GPUs and various things on their motherboard.
Do not do this to $80,000 in GPUs or your surrounding $15,000 computer. Get a couple mcio risers. Lots of cases give you cool ways to still mount more cards like this in the case and keep it clean.
Like why gimp the cards && bake them? there's no reason to do this. If you MUST stack and MUST make it pretty, use water or MaxQ (even these have trouble stacking).
I say this as someone who has done lots of even sillier and dumber things, and will continue to, but I'll do my best to help others not make the same mistakes.
@LLMJunky Like this?
7
3
71
5,419
Harrison Kinsley retweeted
@JasonDClinton @hinapansari “Guys were building a new atomic bomb. Maybe someone should do something about it.
Not us though, we’ll just keep building.”
Not us though, we’ll just keep building.”
3
14
440
39,119
Harrison Kinsley retweeted
We're shipping Asimov 1!
Builders have already received their kits in 🇺🇸 the US, 🇩🇪 Germany, 🇫🇷 France, 🇵🇱 Poland, 🇹🇭 Thailand, and 🇦🇪 the UAE.
We're forming Batch 4 now.
Order now: menlo.ai
Builders have already received their kits in 🇺🇸 the US, 🇩🇪 Germany, 🇫🇷 France, 🇵🇱 Poland, 🇹🇭 Thailand, and 🇦🇪 the UAE.
We're forming Batch 4 now.
Order now: menlo.ai
@asimovinc When does it ship?
9
20
264
11,557
I get endless of these. They always just say the name and price. Or ask me my price for a paid post and dont even tell me about the product, as if that part is irrelevant.
For everyone promoting things this way on all sides, it *is* irrelevant.
For everyone promoting things this way on all sides, it *is* irrelevant.
@eliebakouch I get so many messages to get me to do stuff like that
6
0
29
5,807
The moment i see a video like this i move along. Accts that push this stuff with lazy effort get muted.
You might get a big view count that you can incl in ur next investor email, but people are getting fed up wit this slop.
I already forgot the company in the vid.
You might get a big view count that you can incl in ur next investor email, but people are getting fed up wit this slop.
I already forgot the company in the vid.
i find it really sad that X is becoming a pay to win platform and that even "good" releases are going through these fully slop/fake campaigns to get traction
i'm also pretty sure there is very shady stuff where they are boosting each other's posts with bot farms and more. would here @elonmusk and make sure the algorithm doesn't push their posts, you can very easily cluster that if you look at each release and the accounts amplifying 🙏
(not a dunk on recursive, recently a lot of people are doing that unfo and it's definitely working, posts often get over ~1M views)
4
1
71
5,387
deepseek harness is actually quite good with dsv4.1f 1017
I've been ignoring and avoiding these browser based harnesses, but hosting it locally is a great experience!
anyone tried the zai's zcode with the GLMs? looks very similar.
I've been ignoring and avoiding these browser based harnesses, but hosting it locally is a great experience!
anyone tried the zai's zcode with the GLMs? looks very similar.
17
1
135
8,736
fall recovery is definitely one of the more fun RL policies that I've worked on.
still a ways to go, this one would definitely break a lot of stuff as is, but we'll continue improving! still just a lil baby model.
still a ways to go, this one would definitely break a lot of stuff as is, but we'll continue improving! still just a lil baby model.
I kneel...
training @asimovinc to transfer from lying on the ground (back/belly/side) to a half kneel, with hopes of translating next from a half kneel to full stand.
Trying to find the best way to get up off the ground.
2
4
49
4,089
I kneel...
training @asimovinc to transfer from lying on the ground (back/belly/side) to a half kneel, with hopes of translating next from a half kneel to full stand.
Trying to find the best way to get up off the ground.
training @asimovinc to transfer from lying on the ground (back/belly/side) to a half kneel, with hopes of translating next from a half kneel to full stand.
Trying to find the best way to get up off the ground.
3
2
33
7,761
Harrison Kinsley retweeted
64 concurrent requests. Nearly 7,000 tokens/sec. One machine. 👀
@Sentdex tests Nemotron 120B A12B on the ASUS ExpertCenter Pro ET900N G3.
Learn more: asus.click/ET900N-G3-Soci…
#ASUS #ASUSAgentComputer #ExpertCenterPro #ET900NG3 #GB300 #LocalAI
@Sentdex tests Nemotron 120B A12B on the ASUS ExpertCenter Pro ET900N G3.
Learn more: asus.click/ET900N-G3-Soci…
#ASUS #ASUSAgentComputer #ExpertCenterPro #ET900NG3 #GB300 #LocalAI
dario begged me in dms to not do agent swarms at home but whatever, this is nemotron 120B A12B running 64 concurrents at a bit over 100 tok/sec each for an aggregate of almost 7K tok/sec on a single GB300 (Asus DGX Station)
believe it or not, the agents did NOT magically just t tho. weird.
7
6
52
12,402
The memory and copious dopamine of damp, stand, walk, after you built your Asimov piece by piece, is something you will never forget.
Every great machine begins on a workbench. Now you can build a humanoid robot on yours. docs.menlo.ai
3
2
48
4,747
you can buy a whole dell workstation with an RTX Pro 6K WS inside of it for less than most places will sell you just the pro 6k
42
19
719
65,488
Harrison Kinsley retweeted
DeepSeek V4.1 Flash 1017 scores 45 on the Artificial Analysis Intelligence Index, improving on DeepSeek V4.1 Flash by 6 points, and is the equal #2 open-weights model
@deepseek_ai has released DeepSeek V4.1 Flash 1017, an update on their flagship model DeepSeek V4.1 Flash. On epSeek V4.1 Flash 1017 is priced at $0.30 per 1M input tokens and $1.20 per 1M output tokens during peak hours, with off-peak pricing half of peak rates, the same pricing as DeepSeek V4.1 Flash.
Key results:
➤ DeepSeek V4.1 Flash 1017 makes gains in agentic capabilities, knowledge and scientific reasoning. DeepSeek V4.1 Flash 1017’s Elo rating on GDPval-AA v2.1 is 1746, an increase of 146 points from DeepSeek V4.1 Flash’s rating of 1600. On AA-Briefcase v1.1, its Elo rating increases by 150 points from 1420 to 1570. On AA-Omniscience, its non-hallucination rate is 33%, a 29-point improvement from DeepSeek V4.1 Flash’s 4%, while seeing an improvement in accuracy from 43% to 46%, pushing its overall AA-Omniscience Index score from -5 to +5. On CritPt, DeepSeek V4.1 Flash 1017 scores 24%, up from 14% for DeepSeek V4.1 Flash.
➤ DeepSeek V4.1 Flash 1017 is now the #2 open-weights model alongside GLM-5.3. With a score of 45 on the Artificial Analysis Intelligence Index, DeepSeek V4.1 Flash 1017 is behind MiMo-V2.6-Pro by only 1 point and ahead of other open-source models like Step 5 Preview and Kimi K3 (both at 44).
➤ With a Cost per Task of $0.26, DeepSeek V4.1 Flash 1017 sits in the most attractive quadrant of Capability vs. Cost. With a Cost per Task of $0.26 and Intelligence Index score of 45, DeepSeek V4.1 Flash 1017 compares favorably to GLM-5.3-Flash at $0.28 per task and a score of 42. At the same time, MiMo-V2.6-Pro (46) at $0.13 per task and GPT-6.1 Sol (medium, 48) at $0.21 per task both offer higher capability at a lower Cost per Task.
Additional model details:
➤ Context window: 1M tokens (equivalent to DeepSeek V4.1 Flash)
➤ Pricing: $0.30 per 1M input tokens and $1.20 per 1M output tokens, unchanged from DeepSeek V4.1 Flash. Cache hit price of $0.006 per 1M tokens, a 98% discount
➤ Size: 552B total parameters (16B active)
➤ License: MIT
➤ Providers: Available through DeepSeek’s first-party API, @FireworksAI_HQ , @baseten, @DeepInfra , @wafer_ai , @Cloudflare and others
@deepseek_ai has released DeepSeek V4.1 Flash 1017, an update on their flagship model DeepSeek V4.1 Flash. On epSeek V4.1 Flash 1017 is priced at $0.30 per 1M input tokens and $1.20 per 1M output tokens during peak hours, with off-peak pricing half of peak rates, the same pricing as DeepSeek V4.1 Flash.
Key results:
➤ DeepSeek V4.1 Flash 1017 makes gains in agentic capabilities, knowledge and scientific reasoning. DeepSeek V4.1 Flash 1017’s Elo rating on GDPval-AA v2.1 is 1746, an increase of 146 points from DeepSeek V4.1 Flash’s rating of 1600. On AA-Briefcase v1.1, its Elo rating increases by 150 points from 1420 to 1570. On AA-Omniscience, its non-hallucination rate is 33%, a 29-point improvement from DeepSeek V4.1 Flash’s 4%, while seeing an improvement in accuracy from 43% to 46%, pushing its overall AA-Omniscience Index score from -5 to +5. On CritPt, DeepSeek V4.1 Flash 1017 scores 24%, up from 14% for DeepSeek V4.1 Flash.
➤ DeepSeek V4.1 Flash 1017 is now the #2 open-weights model alongside GLM-5.3. With a score of 45 on the Artificial Analysis Intelligence Index, DeepSeek V4.1 Flash 1017 is behind MiMo-V2.6-Pro by only 1 point and ahead of other open-source models like Step 5 Preview and Kimi K3 (both at 44).
➤ With a Cost per Task of $0.26, DeepSeek V4.1 Flash 1017 sits in the most attractive quadrant of Capability vs. Cost. With a Cost per Task of $0.26 and Intelligence Index score of 45, DeepSeek V4.1 Flash 1017 compares favorably to GLM-5.3-Flash at $0.28 per task and a score of 42. At the same time, MiMo-V2.6-Pro (46) at $0.13 per task and GPT-6.1 Sol (medium, 48) at $0.21 per task both offer higher capability at a lower Cost per Task.
Additional model details:
➤ Context window: 1M tokens (equivalent to DeepSeek V4.1 Flash)
➤ Pricing: $0.30 per 1M input tokens and $1.20 per 1M output tokens, unchanged from DeepSeek V4.1 Flash. Cache hit price of $0.006 per 1M tokens, a 98% discount
➤ Size: 552B total parameters (16B active)
➤ License: MIT
➤ Providers: Available through DeepSeek’s first-party API, @FireworksAI_HQ , @baseten, @DeepInfra , @wafer_ai , @Cloudflare and others
63
111
1,502
225,985
