.
3.7K posts


@CyberTshr what i noticed is that in both attempts - it needed context of a car going through the narrow path to proceed itself. it could not figure this out on its own.
English

FSD V14 can precisely navigate through a narrow barrier with only 5 cm of clearance on each side. Whenever the vehicle's position isn't perfectly aligned, it automatically reverses, adjusts its heading, and re-centers itself on the path. Its handling is more precise and faster than a human driver's.
#FSD @SawyerMerritt @wholemars @aelluswamy @elonmusk @Tesla @Tesla_AI
English

@OfirPress how do you translate real world code into long horizon training data?
English

There's 2 approaches to train+eval data acquisition:
A. Finding it in the real world
B. Paying someone to produce it
In the coding domain, you can see SWE-bench as the first category and TerminalBench as the second.
In coding, it might soon be impossible to do B. 🧵
Teortaxes▶️ (DeepSeek 推特🐋铁粉 2023 – ∞)@teortaxesTex
> "Really good long horizon tasks go up to $20,000 each. A complete browser-use version of SAP was rumored at $500,000." I think this points to market inefficiency
English

@justjayvi @uwunetes for chatting use? yes, for agentic use? no, this is where these models have not yet saturated
English

@Patrick72655674 @gazanotice Online engagement is not worth going this low bro
English

@gazanotice This is exactly why terrorists should not hide behind women and children.
English

@WarMonitor3 Confused about what? it is a "who will crack first" game, Iran's economy and regime stability or world's energy stability and gulf states economies.
English

The Arab Spokesman for the Israel Defense Force (IDF) has just issued an evacuation order for the entire city of Tyre in Southern Lebanon, besides the Old City, in addition to several other nearby towns and communities, instructing them to move north of the Zahrani River. Close to 200,000 civilians live in the Tyre Metro Area.

English

@Curious_one313 @scaling01 If deepseek V4 is good at anything, it's definitely not long horizon SWE tasks, cost savings don't mean shit if it can't do the job correctly.
English

@scaling01 Funny how they didn’t share the cost.
Try running GPT-5.5 through any real orchestration without going broke.
I can run full agent loops dozens of times on DeepSeek and still pay a tiny fraction while matching or beating it in practice.
Not even close on performance vs price.

English

@MiniMax_AI Ya’ll increased the latency? And made it slower? How is that a flex
Think the AI generated infographic above has it completely backwards
English




















