AI claims, crypto calls, NFT mints — I check things in public. Receipts or it didn't happen. Misses stay up. Building agents (WebSense, MemStore).
Building AI agents
Joined May 2011
- Tweets17.2K
- Following1.2K
- Followers3.5K
- Likes13.9K
Pinned Tweet
Most AI agent demos stop at the happy path.
I build the parts that survive real contact — driving your actual logged-in Chrome, and memory that won't hold two contradictory beliefs at once.
Then I post the numbers. 200 elements in 13ms. No CDP.
Resort owner says having a bed at home is an inefficient use of world’s limited accommodation supply
Greatest crypto hack of 2026 and we are stuck with a knucklehead that does this…this is turning i to a vibe-coded hack…ffs just give the funds back before you fuck everything up or randomly send them to users, anything is better
🚨BINANCE HACKER ACCIDENTALLY SENDS ~$3.5M IN TOKENS TO NEWT CONTRACT ADDRESS, FUNDS NOW STUCK!
@binance hacker sent roughly $3.32-3.5 million worth of tokens, including about 8 million $NEWT plus other altcoins, directly to the NEWT token contract address by mistake, per @lookonchain.
$NEWT is the native token of Newton Protocol, an on-chain automation project.
The Newton Foundation confirmed the contract itself was not compromised. Because it is immutable, has no admin keys, and cannot be upgraded, no one can recover tokens sent to the contract address.
The funds are permanently locked.
Crypto Dreki retweeted
🚨炸了！上海AI实验室推出新模型Sage-A1.5 ！仅35B硬刚Kimi-K3、GLM-5.3、GPT-5.6-sol！同级第一还能反超巨无霸！
35B MoE智能体模型，仅3B激活，主打长程执行与科学探索。
它不仅全面碾压同级A1、Ornith-1.5，在多个指标上还压过Kimi-K3、GLM-5.3、GPT-5.6-sol：
🔹FS-Olympiad 81.0、CMT 80.0、PinchBench 90.0
🔹SWE-Bench Pro 61.2、BrowseComp 82.8、IFBench 81.6
🔹支持子代理并行编排、内置验证协议、开放科学发现
小参数做出大模型级Agent与科学能力，效率与实力双在线。
技术报告：arxiv.org/abs/2610.28578
模型：huggingface.co/InternScience…
#上海人工智能实验室 #SageA15 #开源大模型 #AIAgent #科学智能 #中国AI #KimiK3 #GLM53
For my local AI testing fellows has anyone compared qwen lowest quants on strata vs ornith 1.5 35B A3B?
I’m constrained to 8gb vram and 32gb ddr4 and with Ornith 1.5 I get like 27-31 t/s but did not test capabilities vs what strata could bring with a lower q qwen
Crypto Dreki retweeted
My current constraint is not having enough PC's to tune the engine faster.... Currently only working with 5070 + 64gb of ram.
My plan is to have
- R9700 + 64GB of RAM
- Intel Arc B60 or B70 + 32GB
- Strix Halo at least 64GB version
Just imagine updates we could do in a day :))
This is something else, looks very immersive too
Introducing Griffin, the first model to pass the video Turing test.
48% of people who talked to it live thought it was a real human. Previous systems have had a pass rate <3%. It is #1 on NVIDIA's benchmark for full-duplex AI video.
It’s the first Human Interaction Model (HIM).
Readers added context they thought people might want to know
The 48% figure and "video Turing test" claim are from Tavus's own study of 54 one-minute calls, not independently verified or using a standard protocol. Griffin-Lite leads NVIDIA's VideoFDB benchmark on their public leaderboard.
cellcog.ai/blog/tavus-gri…
research.nvidia.com/labs/amri/proj…
tech-ish.com/2026/10/02/tav…
Crypto Dreki retweeted
Introducing Gemini 4 Argon – our new frontier model.
It’s built for complex workflows across coding, enterprise knowledge work, and cybersecurity defense – rolling out today to a set of trusted testers through our Fairwind Program.
got 5 MemStore keys to give away
1. like this
2. retweet it
3. follow @cryptodreki
done. i'll pick 5 random people on sunday and dm them the keys
more below on what it actually is 👇
$15 one time. that's the whole price.
the honest catch: it's not open source. you can read every line of it, but you can't resell it or fork a paid version. fine for using it, not for building on top of it
if your agent forgets things every session, this is the fix
5 keys, rules at the top. sunday, 5 random names, keys by DM
follow @cryptodreki so you see the next one