Thread Writer || Content Creator || Artist || Researcher

Joined June 2023
907 Photos and videos
Pinned Tweet
Sentient has introduced a new framework called CryptoAnalystBench to evaluate Crypto AI Agents. We know AI models are improving rapidly, but grading them in a fast-moving field like crypto is difficult. New information appears constantly, and standard tests often miss the mark. Old benchmarks usually focus on simple retrieval tasks, like checking a token's price. But today, users are asking complex questions that require reasoning, such as why the market is moving a certain way. We need answers that explain the context, not just the raw data. This benchmark was built to fix that evaluation gap. It is a test designed to see how well an AI can handle deep, time-sensitive questions. The goal is to help developers understand what real users actually need. Sentient built this by looking at real data. They started with millions of queries and filtered them down to 198 high-quality questions across 11 categories, including Project Fundamentals and On-Chain Analytics. To grade the answers, they look at four specific factors: ● Relevance (Did it answer the specific prompt?) ● Temporal Relevance (Is the data fresh?) ● Depth (How detailed is the analysis?) ● Consistency (Is it factually correct without contradictions?) The judging is handled by DeepSeek-v3.1 in a blind test to ensure fairness. The results are worth looking at. @SentientAGI open-source model, SERA Kimi K2.5, achieved the highest score, beating out Gemini 3 Pro and GPT-5.2. It also ranked first more often than any other model in head-to-head comparisons. This proves that open-source models are becoming highly competitive with top-tier systems. However, the test also highlighted that factual accuracy is still a challenge for everyone in the space. Hopefully, we see more progress there soon. You can read the full breakdown here: blog.sentient.xyz/posts/cryp…… GitHub: github.com/sentient-agi/Cryp…
4
1
10
4,502
Which planet is this?
4
Why $Zama
$zama is back
2
131
$zama is back
2
3
296
Real 😂
12
Is it worth buying a blue tick on Twitter?
1
31
Guys, how many points do you have in @krakenpro ? Let me know in the comments. And what do you think, how will @inkonchain perform?
1
43
I am back
1
2
27
Big daniel Today chapter last picture #lookism
91
James vs tom. Lookism latest chapter last panel #lookism #latestchapter #highlights
1
2
5
960
Again 888.88
888.88
2
52
Why are the clouds yellow? If anyone knows about this, tell me in the comments. #yellowclouds #sky
1
45
Based Test post @base
Apr 23
This is literally the easiest job ever - Post “Based” - Get likes
27
0xINFO HUB retweeted
I am fully back 🚀 Time to resume bull posting about one of my favorite projects — @inkonchain $INK
I am fully back 🚀 Time to resume bull posting about one of my favorite projects — @inkonchain $INK Earned 14 @nadoHQ points last week 💪 Still got 5 invites left 👀 Who’s ready to run it up?
1
2
12
324
0xINFO HUB retweeted
Guys, which one are you more bullish on, $INK(@inkonchain) or $Base(@base)? Both have exchanges behind them. Share your thoughts 💭 in the comments section 👇.
1
2
6
148
I want this @inkonchain
1
2
18
888.88
2
100
Guys, @inkonchain is giving a hint here to stay as active as possible on @krakenpro. Try to earn as many points as you can. More points = Bigger allocation. Tell me in the comments if you got points or not, and if yes, how many?
Apr 13
Pro tip: activity on @KrakenPro ages well if you’re outside touching grass rn, that’s on you 🫵 get positioned on Kraken Pro wINK wINK
1
1
2
69
A lot of people can’t stop using crypto because of one simple reason. Here you get unexpected airdrops - sometimes $1k, sometimes $10k, sometimes even $100k lottery. In a job everything is fixed, so the excitement dies. Crypto keeps the dopamine alive.
25