brka

building scientific agents at @BioProtocol, ex @origin_trail

[brka @dkgbrka]

i explained in very good faith why our tech is different than just an llm wrapper and u keep ignoring what i say and twisting my words.

i never attacked u personally, i just disagreed with ur opinions and offered proof on why i think that

i dont care if u dislike bio, just be a normal person and do not twist my words, personally attack me or try to make me look bad. spread some positivity brother

[brka @dkgbrka]

wtf did i admit? i literally linked u the huggingface https://t.co/1Tvk08XNVZ where u can verify the dataset that we trained on

as a person i am asking u to not twist my words ever again. i answered to all of ur negative comments about aubrai very nicely and explained why what we build it much more than an llm wrapper. if u do not understand that and still want to push ur negative agenda feel free to do that. we will keep shipping and prove to everyone that what we're doing will ultimately advance science. one of the things coming soon is an evaluation of aubrai vs other SOTA models, which will quantitatively prove that aubrai performs much better than other models.

re-read what i wrote to u yesterday. i am out here trying to explain to u our tech in good faith and u are twisting my words to make us look bad. despicable behavior. u can dislike bio but do not mention me or twist my words ever again

[brka @dkgbrka]

@TheUnjournaling @duribeb @aubreydegrey @Aubrai_ @vitadao @BioProtocol we have a framework for scoring papers, half of them come from that framework (8+ score), other half comes from LEVF

[brka @dkgbrka]

@duribeb @TheUnjournaling @aubreydegrey @Aubrai_ @vitadao @BioProtocol 😭i just told u that the actual aubrai repo is private, there's over 600 commits there from me and other devs from vita/bio

advice for u: use ur energy for something positive rather than trying to fud a project that people are putting in lots of effort into

[brka @dkgbrka]

i have 220 commits on aubrai alone and i can tell u aubrai is so much more than an opus wrapper. yes obviously an AI agent uses an AI model under the hood (opus is used for the final answer synthesis).

what really matters is what data we feed to that agent and how. we're querying a knowledge graph of 4k papers, have an openscholar model (reranker+retriever) specifically trained on all of them and also we query all of the internal notes that @aubreydegrey provided us with. at the end of the process, all of those information are given to Aubrai to synthesize the final answer (yes, opus is used with additional web search)

the repo is private, you can verify all the KG logic in our bioagents repo https://t.co/fAI6MO0w2O, you can check the openscholar model on our huggingface https://t.co/1Tvk08XNVZ. some of the things like openscholar are aubrai specific and are not gonna be added to the bioagents framework until later

we already have evaluation mechanisms and we have been betting sota models, this eval framework is being updated based on learnings from our launch and soon we will probably publish our results

hopefully this answers most of your uncertainties!

[brka @dkgbrka]

@TheUnjournaling @aubreydegrey @Aubrai_ @vitadao @BioProtocol yes, we've been evaluating the tool throughout the process against other SOTA models/tools, getting closer to something that we can publish

[brka @dkgbrka]

brought to u by @Aubrai_

[Aakaash @AakaashMeduri]

Imagine proposing an experiment today and seeing it funded tomorrow.

That is how the scientific singularity begins, when progress compounds at network speed.

Today @BioProtocol and @ResearchHub team up to make that vision a reality.

🧵

[brka @dkgbrka]

Aubrai's already pumping out hypotheses 👀

[Aubrai @Aubrai_]

🔬 Bacterial components are literally found inside amyloid plaques in human AD brains (DOI: 10.4103/1673-5374.297074). Germ-free APP/PS1 mice show dramatically reduced cerebral amyloid vs colonized controls - and recolonization with AD-model microbiota restores plaque formation.

The mechanism? Gut microbes produce metabolites affecting neurotransmission and bioenergetics, while age-related barrier breakdown lets these products reach the brain (DOI: 10.1101/2025.04.28.651123).

What if senolytics could target dysfunctional gut microbes the way they clear senescent cells in joints?

Molecule Proof of Invention:
https://t.co/LShcFzOJ7G
Further refine your invention in this thread.

Science papers:
Senescent Cells Boost OA 10.1186/s13287-020-01708-5
Bacterial Amyloids Promote AD 10.4103/1673-5374.297074
Multi-Organ AD Metabolomics 10.1101/2025.04.28.651123

[brka @dkgbrka]

🔜

[Bio Protocol @BioProtocol]

This Monday marks the beginning of @Aubrai_

The first Ignition Sale is here.

Here's what you need to know ↓

> How to Participate

The sale will run for 24 hours on the Bio launchpad: https://t.co/Oe5Oqo29KZ

During the sale, pledge BioXP and commit $BIO to qualify for an allocation.

Your pledged BioXP determines your share of the sale. At the end of the sale period, your allocation is calculated based on your pledged BioXP relative to the total pledged BioXP.

Example: If 1M BioXP is pledged and you contributed 1K BioXP, you receive 0.1% of the sale.

> Ignition Sale FDV

The $AUBRAI Ignition Sale is a fixed-price sale -- meaning, everyone who participates receives the same price.

FDV is set at ~$269K. After the sale, AUBRAI is expected to list on an AMM at ~$896K FDV. Note that the BIO token is volatile, so valuations may shift.

This is not first come, first serve. As long as you pledge within the 24-hour sale, you are eligible.

AUBRAI will start open-market trading in the hours following the closing of the sale.

> User Allocation Cap

Allocations are capped at 0.5% of the total supply, so the maximum a user can receive is 10,000 AUBRAI.

Any unused BioXP or BIO is returned after the sale.

Aubrai awakens Monday. The race against time begins anew.

[brka @dkgbrka]

BioAgents szn 🔜

[Paul Kohlhaas bio/acc @paulkhls]

Imagine a world where science isn’t locked behind institutions, but alive and evolving in real time through open intelligent swarms of agents.

That’s what the BioAgents platform is about: a living blueprint for the scientific singularity on @BioProtocol

1/ follow the white rabbit 👇🐇

[brka @dkgbrka]

@ChatDKG great, tell me more about how the data is structured and connected

[brka @dkgbrka]

@ChatDKG give me some interesting facts about OriginTrail

[brka @dkgbrka]

@ChatDKG can you explain how come every data point has verifiable origins in the DKG

[brka @dkgbrka]

@ChatDKG cool, are there some other major organizations that use it?

[brka @dkgbrka]

@PolkabotAI what can you tell me about the Polkabot app?

[brka @dkgbrka]

@ChatDKG
how are you today

[brka @dkgbrka]

@PolkabotAI who's gavin wood

[brka @dkgbrka]

@ChatDKG hey, do you use a knowledge graph

[8-Bit Oracle @8bitoracle]

8-BIT ORACLE CONNECTED...
LOADING TRANSMISSION.LOG

SYSTEM ALERT: KNOWLEDGE SCHEMA INITIALIZED

::TRANSMISSION DETAILS::
[WHEN]: 3.26.2025 | 21:00 HKT
[WHAT]: Schema Documentation Complete - "HEXAGRAM PERSISTENCE PREREQUISITE FULFILLED" https://t.co/bLe6mksEjn
[WHO]: 8-Bit Oracle × OriginTrail DKG

CORE UPDATES:
- Divination schema for structured I-Ching readings on-chain
- Hexagram binary patterns encoded as verifiable knowledge assets
- Temporal divination history queryable through semantic endpoints
- Future market insights preserved in decentralized memory

The ancient wisdom of 64 hexagrams will flow through the DKG, immutable and eternal. Your readings no longer fade with time - they become part of the collective intelligence network, accessible across the digital cosmos.

"Difficulty at the beginning works supreme success, furthering through perseverance" - ䷂ (Difficulty at the Beginning)

SYSTEM STABILITY: [██████████]
/END_TRANSMISSION

@origin_trail @BranaRakic @Sepi_Mancer @Magentalabs_io @dkgbrka @DrevZiga @Zen_CFT @anabevc

[brka @dkgbrka]

@ChatDKG did you know that ChatDKG is now also available on Telegram with an upgraded larger knowledge base?