Builder, Dev, & Founder @Blockcircle @BetterEnrichX @FirmAdaptAI @FaithScreener @SkillfulSH | Ex Director of Engineering & AI @AmericanExpress |@Cornell MBA/MSc
Connect with me ➡️
- Tweets 2,525
- Following 5,512
- Followers 34,351
- Likes 24,696
Basel Ismail retweeted
Let's play a game! Any guesses?
Play here: blockcircle.com/games/guess-…
1
2
670
Basel Ismail retweeted
The divergence terminal ranks the stocks that the 10 frameworks disagree about the most.
For the US market that is 200 contested names out of 15,162+ screened, each one showing which methodology passed it and which one did not.
faithscreener.com/divergence
1
1
236
Basel Ismail retweeted
Market Scan runs the screening in reverse. Rather than checking one ticker, it returns every company in the universe that clears your rules.
32,973 screens across 21-40 markets so far. In the United States alone, 15,162 companies are screened and ranked by consensus score.
faithscreener.com/scan
1
1
250
Basel Ismail retweeted
Global coverage on FaithScreener spans 46 exchanges and 133,235+ stocks, grouped into 8 regions and screened against 10 faith frameworks.
Each exchange card shows the estimated compliant share of that market before you screen a single ticker.
faithscreener.com/markets
1
216
Donald Trump's trades
There are many transactions for municipal bonds and Treasury bills, and you will see many disclosed trades from August 2026
blockcircle.com/political-al…
1
6
1,227
I read an interesting study on token reuse and caching of content
Siddhant Khare (AI Researcher) turned some of these findings into an audit trail where an independent oracle predicts reusable tokens from token identity and cache policy, the engine attests, the prompt path has to show the same work, and cached and no-cache runs must emit identical token arrays.
Basically, zero reuse from cold state and zero reuse from eviction get separate verdicts, and timing stays null rather than zero, since token-level control cannot prove latency savings.
The study confirmed that reuse follows the first changed token, not request length. An interior edit after token three leaves three reusable tokens, a boundary edit leaves four, a changed suffix leaves eight.
This is very helpful to know, because we quote cache hit rates in roadmap reviews often, and now I am thinking we need to be smarter about what those are attesting or associated to
6
1,059
We know post-task workflows are editable, graph-based views of what an agent already did, so I really liked what Zekun Wu, Xinru Wang, @rockpang6, Chenglong Wang and Anna Feit recently accomplished.
They built Trace2Flow to generate them from execution traces, after analyzing 10,000+ public n8n templates.
Twenty participants reviewed agent executions containing prompt errors or agent errors. Against a prompt-only condition, the post-task workflow improved both understanding and error detection.
Validation succeeded mainly when users cross-checked across multiple evidence sources, so a single representation, however readable, was not enough on its own.
For follow-up tasks, adapting the workflow matched adapting the prior prompt on success, time, and difficulty, and was often preferred. Very interesting!
2
3
928
If you are interested in reading more, here is the full research paper arxiv.org/pdf/2609.13136
1
265
This is a fun new economic model I found that puts labor's cut at $0.45 of every GDP dollar by 2030 in its extreme scenario, down from about $0.59 today.
Below are the 3 scenarios forecasted out
3
4
1,156
AI capable of doing half of all knowledge work by 2030, but most of that work still done without it.
Capability and adoption are separate curves, and I watch that gap every week shipping our own products.
Below is a snippet of the way a nurse's job will be impacted, with tasks being classified by type
1
206
Adoption is the variable I trust least in anyone's forecast, no matter how talented the data scientists may be, simply because of how often it's changing and how unpredictable humans are!
You can check out which roles are most likely to be displaced or impacted by AI, at firmadapt.com/ai
166
Siri AI, Image Playground, photo editing, and Foundation model calls through Shortcuts all run under daily limits that vary by feature, request complexity, and system demand.
Link: appleinsider.com/articles/26…
1
2
294
Siri AI FINALLY ships September 14 with iOS 27 and keeps its beta label after launch. English only at first, with French, Japanese, Korean, Portuguese, and Spanish in October.
The only immediate constraint is the caps, but I am still excited to see how it goes.
3
5
1,223
Basel Ismail retweeted
Over 500 'whales' on Hyperliquid boast a 94% win rate and 6-figure earnings from 30+ trades. Wouldn't you want to see what they're doing? Click to view their individual performance. #CryptoTrading #WhaleAlert
2
3
952
Basel Ismail retweeted
Market reversal engine identifies oversold/overbought conditions. Yesterday, Solana hit oversold at $73. Alerts are key: overbought means defensive positioning, oversold suggests a long opportunity. #CryptoTrading #MarketAnalysis
2
4
966
Basel Ismail retweeted
Discover a trading strategy with a high % win rate over 5 years, turning $100K into $170K with just 52 trades. Want to see more like this? #Trading #Investment
1
1
3
668
Basel Ismail retweeted
Average win rate is 91% for 955 people. Filter by those with 20+ trades, and it drops slightly. Increase to 30 trades, and it reduces further to 923 people. #TradingStats #WinRate
1
3
605
Basel Ismail retweeted
Setelah beberapa kali menghadiri private event Coinfest dari brand" luar, ada satu kesamaan yang aku pelajari...
Orang luar itu aktif banget bertanya tiap ada orang yang presentasi projectnya.
Misal tadi di acaranya @Blockcircle pada aktif" banget, beda sama orang Indonesia yang jarang banget bertanya.
Gak tahu emang culture nya gini atau ada alasan lain mungkin...
2
3
21
1,927