LogoThread Easy
  • Explorar
  • Criar thread
LogoThread Easy

Seu parceiro completo para threads do Twitter

© 2025 Thread Easy All Rights Reserved.

Explorar

Newest first — browse tweet threads

Keep on to blur preview images; turn off to show them clearly

Be me, think you are very smart, and still get scammed by scammers half my age.

- Land in Hong Kong (still not China)
- Be aware there are lot of scammers trying to sell you electronics product
- Think I have been here many many times, I am scam proof
- Go for a walk
- Run out of battery
- Walk into a store to buy a power bank (this is in a busy tourist area)
- The teen working in the store first shows me regular ones at 30-40 USD.
- She then brings a special one and upsells me this 20,000mAH battery pack
- It has beautiful packaging, looking like an item from a sci-fi.
- Great thing it can be carried on train/airplanes in China (@levelsio, thanks for the reminder)
-  The teen tells me it is 130W, can charge Macbook Pro for 2 times, more like 3 times
- The teen tells me it is twice the price compared to all other batteries as it is Hongkong brand (yeah, now you know where it is going)
- I read the packaging "serving you since 40 years" or something like that.
- I agree to pay the asked price 110 USD (haggling is below my dignity, later I see it should not be above 70-80 USD. Who cares for 30$! Other packs were at 30-40 USD, but I always pay for quality you know).
- I remember that flash of light in her eyes when I agreed to pay. She probably couldn't believe her luck.
- I thank the teen for being such a wonderful salesperson. I say to her I hope she gets enough time to study, and she will be very successful one day (she really upsold me really well)
- She looks down refusing to make eye contact, as if now feeling guilty.
- I charge mobile a bit in the store, the power bank drains rapidly. I think may be it wasn't charge properly (yeah there is such a thing. Stupid me.) 
- Come back to hotel. After charging it to 100% - I now try to charge my Mac. 
-I am dreaming all the freedoms it is going to offer me. Sit in a park, or by the beach and work for hours without worrying about power.
- Then notice, it can not even charge my laptop to 50% when it was supposed to charge it at least 2 times, more likely 3 times.

I read about the brand online. Find nothing!
Try to go to its insta page. Doesn't exist!

I can still go back to the store and argue, but will cost me 1-2 hrs. Not worth. Also, given how dense Honk kong is may not even be able to find it. 

These businesses are predicated on fact that most outsiders won't return to argue. They keep such "special items" hidden from locals (probably).

Not much money, but it hurt me a bit how easily i got scammed. This electronics goods scam is pretty common in this part of the world, I knew it and still fell for it so easily😂.

Be me, think you are very smart, and still get scammed by scammers half my age. - Land in Hong Kong (still not China) - Be aware there are lot of scammers trying to sell you electronics product - Think I have been here many many times, I am scam proof - Go for a walk - Run out of battery - Walk into a store to buy a power bank (this is in a busy tourist area) - The teen working in the store first shows me regular ones at 30-40 USD. - She then brings a special one and upsells me this 20,000mAH battery pack - It has beautiful packaging, looking like an item from a sci-fi. - Great thing it can be carried on train/airplanes in China (@levelsio, thanks for the reminder) - The teen tells me it is 130W, can charge Macbook Pro for 2 times, more like 3 times - The teen tells me it is twice the price compared to all other batteries as it is Hongkong brand (yeah, now you know where it is going) - I read the packaging "serving you since 40 years" or something like that. - I agree to pay the asked price 110 USD (haggling is below my dignity, later I see it should not be above 70-80 USD. Who cares for 30$! Other packs were at 30-40 USD, but I always pay for quality you know). - I remember that flash of light in her eyes when I agreed to pay. She probably couldn't believe her luck. - I thank the teen for being such a wonderful salesperson. I say to her I hope she gets enough time to study, and she will be very successful one day (she really upsold me really well) - She looks down refusing to make eye contact, as if now feeling guilty. - I charge mobile a bit in the store, the power bank drains rapidly. I think may be it wasn't charge properly (yeah there is such a thing. Stupid me.) - Come back to hotel. After charging it to 100% - I now try to charge my Mac. -I am dreaming all the freedoms it is going to offer me. Sit in a park, or by the beach and work for hours without worrying about power. - Then notice, it can not even charge my laptop to 50% when it was supposed to charge it at least 2 times, more likely 3 times. I read about the brand online. Find nothing! Try to go to its insta page. Doesn't exist! I can still go back to the store and argue, but will cost me 1-2 hrs. Not worth. Also, given how dense Honk kong is may not even be able to find it. These businesses are predicated on fact that most outsiders won't return to argue. They keep such "special items" hidden from locals (probably). Not much money, but it hurt me a bit how easily i got scammed. This electronics goods scam is pretty common in this part of the world, I knew it and still fell for it so easily😂.

Artificial Intelligence @amazon, @awscloud Reinforcement Learning, OSS AI, General Purpose Agents, Autonomy All views personal!

avatar for GDP
GDP
Wed Dec 24 00:59:18
Interesting change of strategy by GLM4.7 
compared to Kimi K2 Thinking, DeepSeek V3.2 and MiniMax M2.1

Interleaved thinking between tool calls:
All these models supported interleaved thinking for tool calls, but they clear thinking from the previous turns as can be seen in the first screenshot below.

Preserved thinking in GLM 4.7:
Compare to that GLM 4.7 (for coding endpoints alone)  preserves reasoning from previous turns, as can be seen in screenshot below (note red blocks)

For the other api endpoint, the behaviour is same as before(discard reasoning from previous turns).

This is sure to boost some performance, as the model will have past context. 

As @peakji, advises models need their past thought process to make good decision later. This is anti-context compression, but I guess for coding scenarios may be well worth it.

I wish they made it configurable, so we could see impact ourselves.

Interesting change of strategy by GLM4.7 compared to Kimi K2 Thinking, DeepSeek V3.2 and MiniMax M2.1 Interleaved thinking between tool calls: All these models supported interleaved thinking for tool calls, but they clear thinking from the previous turns as can be seen in the first screenshot below. Preserved thinking in GLM 4.7: Compare to that GLM 4.7 (for coding endpoints alone) preserves reasoning from previous turns, as can be seen in screenshot below (note red blocks) For the other api endpoint, the behaviour is same as before(discard reasoning from previous turns). This is sure to boost some performance, as the model will have past context. As @peakji, advises models need their past thought process to make good decision later. This is anti-context compression, but I guess for coding scenarios may be well worth it. I wish they made it configurable, so we could see impact ourselves.

Artificial Intelligence @amazon, @awscloud Reinforcement Learning, OSS AI, General Purpose Agents, Autonomy All views personal!

avatar for GDP
GDP
Tue Dec 23 06:35:49
Just wow - "wild efficiency gains" with Thinking Machine's approach 🚀🚀🚀

"Post-train with MOPD: We adopted On-Policy-Distillation from Thinking Machine to merge multiple RL models, and the efficiency gains were wild. 

We matched the teacher model's performance using less than 1/50th the compute of a standard SFT+RL pipeline. 

There’s a clear path here for a self-reinforcing loop where the student evolves into a stronger teacher."

Just wow - "wild efficiency gains" with Thinking Machine's approach 🚀🚀🚀 "Post-train with MOPD: We adopted On-Policy-Distillation from Thinking Machine to merge multiple RL models, and the efficiency gains were wild. We matched the teacher model's performance using less than 1/50th the compute of a standard SFT+RL pipeline. There’s a clear path here for a self-reinforcing loop where the student evolves into a stronger teacher."

Artificial Intelligence @amazon, @awscloud Reinforcement Learning, OSS AI, General Purpose Agents, Autonomy All views personal!

avatar for GDP
GDP
Sun Dec 21 19:24:10
This is going to be multi-billion revenue toy. Movies could become a side income for Disney, if this  happens.

This is going to be multi-billion revenue toy. Movies could become a side income for Disney, if this happens.

Artificial Intelligence @amazon, @awscloud Reinforcement Learning, OSS AI, General Purpose Agents, Autonomy All views personal!

avatar for GDP
GDP
Sun Dec 21 10:05:10
Eliezer Yudkowsky, time to move to India.

What will save us from skynet AGI is a power outage!

Incompetent power grid people, will put AGI in its places.

That means India is very safe from Skynet AGI.

Eliezer Yudkowsky, time to move to India. What will save us from skynet AGI is a power outage! Incompetent power grid people, will put AGI in its places. That means India is very safe from Skynet AGI.

Artificial Intelligence @amazon, @awscloud Reinforcement Learning, OSS AI, General Purpose Agents, Autonomy All views personal!

avatar for GDP
GDP
Sun Dec 21 09:39:08
Congrats Pieter Abbeel @pabbeel - Amazon's new AGI Head.

TPOT needs no introduction to him. He is a pioneer in Robotics and DeepRL. Has 231K citations.

He is Prof UC Berkeley; has been an Amazon Distinguished Scientist / VP / Scholar since 2024 focussed on Advancing AI and Robotics.

He has advised legendary doctoral students like Chelsea Finn (@chelseabfinn) of Physical Intelligence and John Schulman (@johnschulman2) - ex OpenAI, ex Anthropic and now co-founder of Thinking Machines.

Looking forward.

Congrats Pieter Abbeel @pabbeel - Amazon's new AGI Head. TPOT needs no introduction to him. He is a pioneer in Robotics and DeepRL. Has 231K citations. He is Prof UC Berkeley; has been an Amazon Distinguished Scientist / VP / Scholar since 2024 focussed on Advancing AI and Robotics. He has advised legendary doctoral students like Chelsea Finn (@chelseabfinn) of Physical Intelligence and John Schulman (@johnschulman2) - ex OpenAI, ex Anthropic and now co-founder of Thinking Machines. Looking forward.

Artificial Intelligence @amazon, @awscloud Reinforcement Learning, OSS AI, General Purpose Agents, Autonomy All views personal!

avatar for GDP
GDP
Wed Dec 17 18:05:48
  • Previous
  • 1
  • 2
  • 3
  • More pages
  • 36
  • 37
  • Next