Activity
Mon
Wed
Fri
Sun
Oct
Nov
Dec
Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
Sep
What is this?
Less
More
Jens AI Community

1.5k members • Free

14 contributions to Clief Notes
Jake is a genius but...
I missed the most important video of all. The one which I should've watched first. How impressed and out of touch I am that I enrolled in March this year and just now (almost 6 months later) I watched THE most important video to start ICM (Interpretable Context Methodology: Folder Structure as Agent Architecture). Shame on me. I want to thank Jake's brain and his ICM system, he is so many light years away from a normal human being that having a translator like Jose Quiles from iScaleLabs video in: Classroom "The Foundation" WATCH THIS FIRST Break down from @iscalelabs ! Is the ELI5 of ICM If you are reading this and have any questions regarding ICM watch this video and practice. I attached the transcript of both videos, Jake's and Jose Quiles. Have fun and keep impressing yourself with Jake's methodology! Happy coding!
Jake is a genius but...
e-mail filtering assistant
Hi everybody, I am working on this project to filter out my e-mail accounts. I have several emails from work and personal. Every week I receive around 500 emails. A real pain in ^#$% I have been working on an app with local model so my info doesn't leave my local hardware (I don't like sharing my nor my companies info with external models). Instead of working around figuring out what works and what doesn't work I decided to ask the community. My local setup and some workaround I found working with Mail app is that you can use it with a cron on the background to filter by category any number of email accounts including gmail, yahoo, outlook and imap accounts. So far I have accomplished the app to flag important emails and the ones it can't solve by itself. I have managed to reduce my emails from +13,000 to only a couple of dozen a week. It has its hiccups and the filtering is not perfect which leads me to ask the community if someone has a similar challenge and how did you resolved it. There are some solutions out there but the config process is really tedious with less than optimal results, just as the cron I configured. Any ideas or suggestions will be very appreciated (don't mind my grammar, not my native tongue) :)
1 like • 15d
@Mofedul Alam Joy Thank you! Totally agreed on the headers and that is one of the methods I used to downsize the amount of emails. I run a deterministic .py on the background and what is not totally clear, for instance, banking and any financial institutions needs to be sorted out. Also the renewal services which some are direct messages to be processed. Awesome idea to add all the sent emails to a clean list. That I will implement o Monday. Thanks again @Mofedul Alam Joy !
1 like • 14d
@Mofedul Alam Joy I will. Thanks!
My research outside ICM
If anyone is curious about Bias in data and LLMs I do alot of work in that space as well. I have a new paper in collaboration with Edinburgh University Neuropolitics Research Lab in preprint ! Check it out here ! Ideological Distortions in the Digital Public Sphere: An Audit of Authoritarianism in Commercial LLMs by Constantine George Kyritsopoulos, Adam B. Moore, Laura Cram, Jake Van Clief, Clare Llewellyn :: SSRN
0 likes • 18d
I remember when Google started creating diverse gender and mixed ethnic figures and refused to create even current famous people. One of the best/worse ones was the image of the founding fathers all where different ethnicity and gender and couldn't render anything real. I have to say in my personal opinion, Grok is the least biased of them all and I give credit on their way of training the model in "facts only", in quotations because, for example, history is taught by humans and from the winning side. Although I usually use a Chinese model when in doubt regarding western issues. They tend to be less biased. I don't think the psychometrical tests will work if only answers are viewed and the LLM's are trained in fiction or sentiments and not facts. Good read and starting point to stand the fact about the limitations of the current methods of programing the models.
A bit late after learning
Hey Clief Notes fam! 👋 I'm JP and I'm stoked to be here. 🙋 A little about me: Like a kid learning AI and getting to know the community. Not a systems engineer. I thought I was going to work less and have a better family time with AI ... guess what? It is the contrary endless nights, and the worse part of it is that I am loving it (not my wife)😂 🎯 My current goal: Master ICM process and all that it means. 💪 What I'm currently building/working on: Gosh, I have a few going on and all are work in progress from chaotic programing to a ICM and 60-30-10 @Jake Van Clief method. My ICM implementation method isn't perfect, mainly because I did'n know it existed at first so I started with a similar concept of the 60-30-10 concept. Here is a really brief explanation of my 4 most time consuming ones I am currently working on: 1. PBX Call analytics with voice recognition, sentiment, psychology, merit/demerit, order tracking among others, all locally processed, this project is a personal challenge processing all locally on-line-on-time and historical calls with a dashboard. 2. Camila agent; a Whatsapp agent with 4 sub-agents, a real challenge and headache. Trying to fix lots of issues and transforming it to the ICM method with an icm-audit .py a.- Sales (Camila) analyzes the conversation and gives a score for sales and notifies the sales team via telegram, sounds sexy until the agent starts giving the wrong answers even though its guard rails. b.- Graphic Arts processing (Majo) 100% deterministic process for large format print jobs, process the links for +600Mgb data files, analyzes them, process them for the print job and sends telegram notification and email with the download links to sales and production for printing while notifying the customer, c.- Traffic for internal sales support generates custom presentations and quotations for the sales team, d.-Accounting (Luis) for procurement invoice reception and processing against PO's and notifies auditors, accounting and supplier if it is accepted or denied.
4
0
COMMUNITY COMPETITION MEGATHREAD 🏆
📣 Update (9/7/26): Comp 12 Rules Added --- What this thread is: I’ve seen a number of people either not know competitions are going on or know how to find them so here is a quick single source thread where you can get to them all. I’ll keep this updated as new challenges are released. See the current challenge as well as all past challenges below! Format for community challenges: https://www.skool.com/cliefnotes/how-competitions-work-from-now-on?p=84912d60 Weekly leaderboard competition: Alongside the biweekly challenges, there is a weekly leaderboard challenge where the member at the top of the 7 day rolling leaderboard wins a free upgrade to their membership. This winner is announced on Mondays typically. @Joshua Hubbard has gone another step further and created a thread linking to all competition entries. Link here: https://www.skool.com/cliefnotes/the-competition-archive-every-entry-every-week-every-link?p=bd752987 --- Current Challenge: Competition 12 The Challenge 💪 - Build an auditor that reviews something against a defined standard. Rules & submissions 🗒️ - https://www.skool.com/cliefnotes/weekly-comp-11-the-auditor?p=1d0eb170 Good luck everyone! --- Past Challenges Competition 1 The Challenge 💪 - Build a brand voice document for a pet grooming business Rules & submissions 🗒️ - https://www.skool.com/cliefnotes/first-ever-weekly-competition-is-live?p=90f50bf9 Results🏆 - https://www.skool.com/cliefnotes/weekly-comp-1-winner-ian-barriopedro?p=0501b57e Competition 2 The Challenge 💪 - Build an artifact of a financial advisor client intake system
1 like • 18d
awesome @Don Roy ! I see I lost a few good opportunities. Will be more involved sharing my projects!
1-10 of 14
Jose Guerrero
3
24 points to level up
@jose-guerrero-3580
From Guatemala, getting into new adventures

Active 3d ago
Joined Mar 16, 2026
Guatemala
Powered by