Benjamin Han
@benjaminhan.sigmoid.social.ap.brid.gy
Husband, father, runner, German learner, piano player. A curious soul living in #PacificNorthwest. Working on Knowledge + #ML + #AgenticAI at .
#Running […]
🌉 bridged from ⁂ https://sigmoid.social/@BenjaminHan, follow @ap.brid.gy to interact
#Running […]
🌉 bridged from ⁂ https://sigmoid.social/@BenjaminHan, follow @ap.brid.gy to interact
November #running recap: compared to October I have 171.3 → 222.1mi (231.7 w/ walking), EG 3,384 → 6,113ft, 28 → 35hr running.
This was up significantly from October (including EG), and met both of my monthly quota 200mi and yearly goal 2,400mi (one month […]
[Original post on sigmoid.social]
This was up significantly from October (including EG), and met both of my monthly quota 200mi and yearly goal 2,400mi (one month […]
[Original post on sigmoid.social]
December 1, 2025 at 5:31 PM
November #running recap: compared to October I have 171.3 → 222.1mi (231.7 w/ walking), EG 3,384 → 6,113ft, 28 → 35hr running.
This was up significantly from October (including EG), and met both of my monthly quota 200mi and yearly goal 2,400mi (one month […]
[Original post on sigmoid.social]
This was up significantly from October (including EG), and met both of my monthly quota 200mi and yearly goal 2,400mi (one month […]
[Original post on sigmoid.social]
1/
This #longrunsunday I ran my #72 #halfmarathon and hit my 2,400mi yearly goal a whole month early (2,404mi)! Route was Carnation -> Fall City and back (16.09mi, moving time 2:43:48).
It was a brisk and misty morning -- winter is coming to #pacificnorthwest!
#photos #running #pnw
This #longrunsunday I ran my #72 #halfmarathon and hit my 2,400mi yearly goal a whole month early (2,404mi)! Route was Carnation -> Fall City and back (16.09mi, moving time 2:43:48).
It was a brisk and misty morning -- winter is coming to #pacificnorthwest!
#photos #running #pnw
December 1, 2025 at 2:00 AM
1/
This #longrunsunday I ran my #72 #halfmarathon and hit my 2,400mi yearly goal a whole month early (2,404mi)! Route was Carnation -> Fall City and back (16.09mi, moving time 2:43:48).
It was a brisk and misty morning -- winter is coming to #pacificnorthwest!
#photos #running #pnw
This #longrunsunday I ran my #72 #halfmarathon and hit my 2,400mi yearly goal a whole month early (2,404mi)! Route was Carnation -> Fall City and back (16.09mi, moving time 2:43:48).
It was a brisk and misty morning -- winter is coming to #pacificnorthwest!
#photos #running #pnw
#parkrun #turkeytrot today! A record 270+ runners showed up! I ran a 22:09 — not a PR, but felt strong and in control the whole way.
On this #thanksgiving: huge thanks to Parkrun and all the volunteers for creating a wonderful community where anyone can join […]
[Original post on sigmoid.social]
On this #thanksgiving: huge thanks to Parkrun and all the volunteers for creating a wonderful community where anyone can join […]
[Original post on sigmoid.social]
November 27, 2025 at 10:50 PM
#parkrun #turkeytrot today! A record 270+ runners showed up! I ran a 22:09 — not a PR, but felt strong and in control the whole way.
On this #thanksgiving: huge thanks to Parkrun and all the volunteers for creating a wonderful community where anyone can join […]
[Original post on sigmoid.social]
On this #thanksgiving: huge thanks to Parkrun and all the volunteers for creating a wonderful community where anyone can join […]
[Original post on sigmoid.social]
Not very exciting, but this #longrunsunday I just ran my 71st half marathon on #dreadmill: 14.28mi pace 9’53”/mi, with CD in the last mi. ~90% easy.
#running
#running
November 24, 2025 at 3:06 AM
Not very exciting, but this #longrunsunday I just ran my 71st half marathon on #dreadmill: 14.28mi pace 9’53”/mi, with CD in the last mi. ~90% easy.
#running
#running
November 19, 2025 at 5:36 AM
November 16, 2025 at 4:44 AM
1/
This #longrunsunday I went with my 9th 50K run (since 2024): run from #redmond Library all the way to my office building in #Seattle! 31.12mi, moving time 5:06:22 pace 9’51"/mi.
#running
This #longrunsunday I went with my 9th 50K run (since 2024): run from #redmond Library all the way to my office building in #Seattle! 31.12mi, moving time 5:06:22 pace 9’51"/mi.
#running
November 10, 2025 at 5:31 AM
1/
This #longrunsunday I went with my 9th 50K run (since 2024): run from #redmond Library all the way to my office building in #Seattle! 31.12mi, moving time 5:06:22 pace 9’51"/mi.
#running
This #longrunsunday I went with my 9th 50K run (since 2024): run from #redmond Library all the way to my office building in #Seattle! 31.12mi, moving time 5:06:22 pace 9’51"/mi.
#running
2/n
I asked how it thinks about my run today. I actually fat-fingered the question (was still standing on my treadmill), but it understood me fine.
Then it proceeded to give me a letter grade: a A-minus! I think I got dinged for not raising the treadmill 0 […]
[Original post on sigmoid.social]
I asked how it thinks about my run today. I actually fat-fingered the question (was still standing on my treadmill), but it understood me fine.
Then it proceeded to give me a letter grade: a A-minus! I think I got dinged for not raising the treadmill 0 […]
[Original post on sigmoid.social]
August 20, 2025 at 5:13 AM
2/n
I asked how it thinks about my run today. I actually fat-fingered the question (was still standing on my treadmill), but it understood me fine.
Then it proceeded to give me a letter grade: a A-minus! I think I got dinged for not raising the treadmill 0 […]
[Original post on sigmoid.social]
I asked how it thinks about my run today. I actually fat-fingered the question (was still standing on my treadmill), but it understood me fine.
Then it proceeded to give me a letter grade: a A-minus! I think I got dinged for not raising the treadmill 0 […]
[Original post on sigmoid.social]
2/
For my upcoming races, I decided I’m going all in for my #chatgpt experiment: I’m going to use it as my personal coach for the entire preparation! I already told it the race name and my last Sunday’s race, and then uploaded my run stats today. It basically […]
[Original post on sigmoid.social]
For my upcoming races, I decided I’m going all in for my #chatgpt experiment: I’m going to use it as my personal coach for the entire preparation! I already told it the race name and my last Sunday’s race, and then uploaded my run stats today. It basically […]
[Original post on sigmoid.social]
August 17, 2025 at 11:05 PM
2/
For my upcoming races, I decided I’m going all in for my #chatgpt experiment: I’m going to use it as my personal coach for the entire preparation! I already told it the race name and my last Sunday’s race, and then uploaded my run stats today. It basically […]
[Original post on sigmoid.social]
For my upcoming races, I decided I’m going all in for my #chatgpt experiment: I’m going to use it as my personal coach for the entire preparation! I already told it the race name and my last Sunday’s race, and then uploaded my run stats today. It basically […]
[Original post on sigmoid.social]
August 13, 2025 at 9:46 PM
4/n
It was so bad that I was actually having doubt if I could finish. Well, I persisted, and tried my best to have a strong-ish finish — at least about the same pace as how I started.
I was really disappointed with the result, but after a steak dinner, I […]
[Original post on sigmoid.social]
It was so bad that I was actually having doubt if I could finish. Well, I persisted, and tried my best to have a strong-ish finish — at least about the same pace as how I started.
I was really disappointed with the result, but after a steak dinner, I […]
[Original post on sigmoid.social]
August 11, 2025 at 4:08 AM
4/n
It was so bad that I was actually having doubt if I could finish. Well, I persisted, and tried my best to have a strong-ish finish — at least about the same pace as how I started.
I was really disappointed with the result, but after a steak dinner, I […]
[Original post on sigmoid.social]
It was so bad that I was actually having doubt if I could finish. Well, I persisted, and tried my best to have a strong-ish finish — at least about the same pace as how I started.
I was really disappointed with the result, but after a steak dinner, I […]
[Original post on sigmoid.social]
2/
First, weather. We started out from Snoqualmie Pass at ~62F (16.7C) and finished at North Bend at ~80F (26.7C). There's actually a heat advisory out today (see picture) 🫠. Last year at North Bend around the same time the temp was ~68F (20C).
I grabbed […]
[Original post on sigmoid.social]
First, weather. We started out from Snoqualmie Pass at ~62F (16.7C) and finished at North Bend at ~80F (26.7C). There's actually a heat advisory out today (see picture) 🫠. Last year at North Bend around the same time the temp was ~68F (20C).
I grabbed […]
[Original post on sigmoid.social]
August 11, 2025 at 3:44 AM
2/
First, weather. We started out from Snoqualmie Pass at ~62F (16.7C) and finished at North Bend at ~80F (26.7C). There's actually a heat advisory out today (see picture) 🫠. Last year at North Bend around the same time the temp was ~68F (20C).
I grabbed […]
[Original post on sigmoid.social]
First, weather. We started out from Snoqualmie Pass at ~62F (16.7C) and finished at North Bend at ~80F (26.7C). There's actually a heat advisory out today (see picture) 🫠. Last year at North Bend around the same time the temp was ~68F (20C).
I grabbed […]
[Original post on sigmoid.social]
2/n
What did my #chatgpt coach say about my run today? "Use this run as your confidence benchmark — you’re in race-ready shape.”
See picture for details!
#running #marathon #race #tunnelvision
What did my #chatgpt coach say about my run today? "Use this run as your confidence benchmark — you’re in race-ready shape.”
See picture for details!
#running #marathon #race #tunnelvision
August 3, 2025 at 11:23 PM
2/n
What did my #chatgpt coach say about my run today? "Use this run as your confidence benchmark — you’re in race-ready shape.”
See picture for details!
#running #marathon #race #tunnelvision
What did my #chatgpt coach say about my run today? "Use this run as your confidence benchmark — you’re in race-ready shape.”
See picture for details!
#running #marathon #race #tunnelvision
For my #tunnelvision #marathon #race (https://www.tunnelmarathon.com) on August 10, I asked #chatgpt for the course profile (picture 1), gave my usual weekly training routine, and then asked for a last-week training plan (picture 2). It asked me if I need a […]
[Original post on sigmoid.social]
[Original post on sigmoid.social]
August 3, 2025 at 3:07 AM
For my #tunnelvision #marathon #race (https://www.tunnelmarathon.com) on August 10, I asked #chatgpt for the course profile (picture 1), gave my usual weekly training routine, and then asked for a last-week training plan (picture 2). It asked me if I need a […]
[Original post on sigmoid.social]
[Original post on sigmoid.social]
3/n
- They showed small interventions (like role clarification and verification steps) led to 9–16% improvements (picture).
MAST moves MAS development closer to science and engineering rather than guesswork.
Paper: https://arxiv.org/abs/2503.13657
#llm #ai #genai #mas #agenticai #paper
- They showed small interventions (like role clarification and verification steps) led to 9–16% improvements (picture).
MAST moves MAS development closer to science and engineering rather than guesswork.
Paper: https://arxiv.org/abs/2503.13657
#llm #ai #genai #mas #agenticai #paper
August 2, 2025 at 9:51 PM
3/n
- They showed small interventions (like role clarification and verification steps) led to 9–16% improvements (picture).
MAST moves MAS development closer to science and engineering rather than guesswork.
Paper: https://arxiv.org/abs/2503.13657
#llm #ai #genai #mas #agenticai #paper
- They showed small interventions (like role clarification and verification steps) led to 9–16% improvements (picture).
MAST moves MAS development closer to science and engineering rather than guesswork.
Paper: https://arxiv.org/abs/2503.13657
#llm #ai #genai #mas #agenticai #paper
2/
- They identified 14 failure modes across 3 categories (picture 1):
- Specification issues (poor prompt design, rigid turn-taking): 41.77%
- Inter-agent misalignment (reasoning-action mismatches): 36.94%
- Task verification failures (premature termination) […]
[Original post on sigmoid.social]
- They identified 14 failure modes across 3 categories (picture 1):
- Specification issues (poor prompt design, rigid turn-taking): 41.77%
- Inter-agent misalignment (reasoning-action mismatches): 36.94%
- Task verification failures (premature termination) […]
[Original post on sigmoid.social]
August 2, 2025 at 9:51 PM
2/
- They identified 14 failure modes across 3 categories (picture 1):
- Specification issues (poor prompt design, rigid turn-taking): 41.77%
- Inter-agent misalignment (reasoning-action mismatches): 36.94%
- Task verification failures (premature termination) […]
[Original post on sigmoid.social]
- They identified 14 failure modes across 3 categories (picture 1):
- Specification issues (poor prompt design, rigid turn-taking): 41.77%
- Inter-agent misalignment (reasoning-action mismatches): 36.94%
- Task verification failures (premature termination) […]
[Original post on sigmoid.social]
1/
Multi-agent LLM systems (MAS) promise enhanced capabilities through collaboration, but often fail to vastly outperform single-agent setups. What are their failure modes and how do we mitigate them?
A recent paper from UC Berkeley introduces MAST […]
[Original post on sigmoid.social]
Multi-agent LLM systems (MAS) promise enhanced capabilities through collaboration, but often fail to vastly outperform single-agent setups. What are their failure modes and how do we mitigate them?
A recent paper from UC Berkeley introduces MAST […]
[Original post on sigmoid.social]
August 2, 2025 at 9:49 PM
1/
Multi-agent LLM systems (MAS) promise enhanced capabilities through collaboration, but often fail to vastly outperform single-agent setups. What are their failure modes and how do we mitigate them?
A recent paper from UC Berkeley introduces MAST […]
[Original post on sigmoid.social]
Multi-agent LLM systems (MAS) promise enhanced capabilities through collaboration, but often fail to vastly outperform single-agent setups. What are their failure modes and how do we mitigate them?
A recent paper from UC Berkeley introduces MAST […]
[Original post on sigmoid.social]
Top and bottom 40 occupations that are affected by hashtag#genAI, according to MSR paper:
"We find the highest AI applicability scores for knowledge work occupation groups such as computer and mathematical, and office and administrative support, as well as […]
[Original post on sigmoid.social]
"We find the highest AI applicability scores for knowledge work occupation groups such as computer and mathematical, and office and administrative support, as well as […]
[Original post on sigmoid.social]
August 1, 2025 at 6:34 PM
Top and bottom 40 occupations that are affected by hashtag#genAI, according to MSR paper:
"We find the highest AI applicability scores for knowledge work occupation groups such as computer and mathematical, and office and administrative support, as well as […]
[Original post on sigmoid.social]
"We find the highest AI applicability scores for knowledge work occupation groups such as computer and mathematical, and office and administrative support, as well as […]
[Original post on sigmoid.social]
July #running recap: compared to June I have 245.9 → 225.4mi (240.0 w/ walking), EG 3,728 → 4,419ft, 37 → 34hr running.
This was down from June, but still 25.4mi above my monthly 200mi goal. During the month I ran my 27th marathon and 60th half since 2022. My […]
[Original post on sigmoid.social]
This was down from June, but still 25.4mi above my monthly 200mi goal. During the month I ran my 27th marathon and 60th half since 2022. My […]
[Original post on sigmoid.social]
August 1, 2025 at 3:42 PM
July #running recap: compared to June I have 245.9 → 225.4mi (240.0 w/ walking), EG 3,728 → 4,419ft, 37 → 34hr running.
This was down from June, but still 25.4mi above my monthly 200mi goal. During the month I ran my 27th marathon and 60th half since 2022. My […]
[Original post on sigmoid.social]
This was down from June, but still 25.4mi above my monthly 200mi goal. During the month I ran my 27th marathon and 60th half since 2022. My […]
[Original post on sigmoid.social]
2/n
Notes:
* Still quite a room to reach target pace 7’37”/mi to get under 3:20 finish time for BQ!
* Bottleneck will be on quads and glutes (feeling it even w/ half).
* Got jumbo-sized ice cream at Dairy Freeze after run, and the random person I picked to […]
[Original post on sigmoid.social]
Notes:
* Still quite a room to reach target pace 7’37”/mi to get under 3:20 finish time for BQ!
* Bottleneck will be on quads and glutes (feeling it even w/ half).
* Got jumbo-sized ice cream at Dairy Freeze after run, and the random person I picked to […]
[Original post on sigmoid.social]
July 20, 2025 at 11:36 PM
2/n
Notes:
* Still quite a room to reach target pace 7’37”/mi to get under 3:20 finish time for BQ!
* Bottleneck will be on quads and glutes (feeling it even w/ half).
* Got jumbo-sized ice cream at Dairy Freeze after run, and the random person I picked to […]
[Original post on sigmoid.social]
Notes:
* Still quite a room to reach target pace 7’37”/mi to get under 3:20 finish time for BQ!
* Bottleneck will be on quads and glutes (feeling it even w/ half).
* Got jumbo-sized ice cream at Dairy Freeze after run, and the random person I picked to […]
[Original post on sigmoid.social]
2/
* The evolution of ARC benchmarks
- ARC-1: Prior-free reasoning tasks
- ARC-2: Compositional reasoning tasks
- ARC-3: Reasoning with interactive agency
* Intelligence implementation requires two components
- Abstraction acquisition
- Abstraction recomination (application)
#AGI #ai #talk
* The evolution of ARC benchmarks
- ARC-1: Prior-free reasoning tasks
- ARC-2: Compositional reasoning tasks
- ARC-3: Reasoning with interactive agency
* Intelligence implementation requires two components
- Abstraction acquisition
- Abstraction recomination (application)
#AGI #ai #talk
July 7, 2025 at 6:13 PM