@Maryse Montagne Block themselves had a really good introduction and writeup on Berd found here....https://block.xyz/inside/designing-ai-with-character-what-we-learned-building-berd it explains a lot more of their thinking. I believe the design is very chat like, as simple as sending a txt message really, and be accessible for Non-IT people and early techie people too (with some guidance from External IT). Given the open-source nature of Berd and the Apache2 licensing someone could whitelabel it as well with some "Starters out of the box". I like the idea of the philosophy that Design should also make tech accessible...with a simple interface. e.g. Their out of the box animated AI agents are called Gloopies! <-- maybe its a play on Groupies?!? New Agents themselves can be bundled up as Zip files and shared with others....which makes it easy to "Install and provide a new agent". Skills support is helpful as well. This feels to me like a slimmed down harness that is flexible....allows users to leverage a variety of AI models (e.g. Maybe a specific Video Creation Agent!! or Good Image Creation agent, Voice overs, etc) that could eventually turn into something like an open-source flavor of Marblism. I believe on the commercial side you see something similar with the new Grok Bot - but without the $200/month lock-in subscription to Grok. It's good to train people with as well.... 1) Let's look at a prompt for an Agent activity 2) Let's look at a skill - what that reads like. 3) Let's look at what happens when you run a local agent versus a cloud hosted agent. 4) The price token calculator for each chat is helpful as well. An External IT support person (Paid I'm thinking) would want to make sure API setups are sound, token reduction strategies are in place on agents, and maybe also setup a "Credential Broker" to sit in the middle to obfuscate the real API keys in a service <-- versus putting the highly important API keys into an end user app like this. This type of value add service + providing some additional AI Bots and Skills by an External IT support person would be very valuable I think.

One comparison to watermarks (could they also become malicious?) are malicious PDF files...where attackers hide executable code within PDF documents (here is a good read - albeit long on how that happens https://www.sentinelone.com/blog/malicious-pdfs-revealing-techniques-behind-attacks/) . It's a little extra unintended cargo. There are companies that created solutions to fight this for PDFs ....and I actually believe before a PDF gets loaded into an AI program or filestore they should be evaluated. The process - before loading a pdf the system creates a separate isolated sandbox, introspects the PDF, strips out "unintended scripts and threats", recreates the document, and then saves it in its cleansed state. I'm sure there will be a business for non-AI code programs to cleanse through the watermarks/Cryptographic signatures and produce a "cleansed" output saying it is original. Whether the compute cost warrants the outcome...that's for individuals to decide. I actually like the idea of a cryptographic signature showing provenance of content myself ...https://drainpipe.io/knowledge-base/what-is-provenance-in-ai-generated-content/ . That is one reason for my side project to create truthinresearch.com (*Not live yet). And it may help for research/pulbished papers that cite other pieces of work for all citations to be cryptographically signed (https://www.bbc.com/news/articles/cdj98vrzpyvo). Same for attorney's referencing other cases and laws in court. (https://www.msba.org/site/site/content/News-and-Publications/News/General-News/Massachusetts_Lawyer-Sanctioned_for_AI_Generated-Fictitious_Cases.aspx)

I found this YouTube video and explanation really helpful understanding the approach. Please share the best explanations you've seen too. https://youtu.be/YtSq6d_jUWA?is=bPh8Q-nuX_XEbfXj

I am trying to come up with the design (and getting educated along the way) to migrate a SLACK environment into a Buzz by Block environment. This would create steps to migrate and convert an enterprise/middle size company user of SLACK and load all the historical information including users, communities, DMs, and other content into a Buzz application. The idea behind this is this process will help enterprises move away from SLACK (reduce their per user licensing) into the open source Buzz environment. The potential here is to save millions per organization. The ideal approach is to create and deploy an agent that will eventually shepherd clients through the process...reducing adoption friction. After that....I'd look to do a DISCORD migration environment. And maybe then TEAMS migration.

This makes me think of what banks in Australia came up with to help validate KYC - know your customer. They came up with a balanced scorecard that point scored various data proof items and to take someone on as a customer they had to achieve a certain point level. This is one thing that I like about the signed public and private keys....theoretically it will be more difficult to spoof identities.