Eli Yune

Co-Founder, CEO @balkeumlabs |
Move atoms, not just electrons

[Eli Yune @eli_yune31608]

Federated Learning locks you into one architecture.
Federated Distillation promises heterogeneity, but trades weight communication for proxy dataset bottlenecks and broken logit consensus on Non-IID data.
​Federated PEFT (LoRA) remains the real pragmatic winner.

[Eli Yune @eli_yune31608]

@BurnishedRows Or just promote the first guy to do so, and fire the rest :o

[Andrej Karpathy @karpathy]

Agency > Intelligence

I had this intuitively wrong for decades, I think due to a pervasive cultural veneration of intelligence, various entertainment/media, obsession with IQ etc. Agency is significantly more powerful and significantly more scarce. Are you hiring for agency? Are we educating for agency? Are you acting as if you had 10X agency?

Grok explanation is ~close:

“Agency, as a personality trait, refers to an individual's capacity to take initiative, make decisions, and exert control over their actions and environment. It’s about being proactive rather than reactive—someone with high agency doesn’t just let life happen to them; they shape it. Think of it as a blend of self-efficacy, determination, and a sense of ownership over one’s path.

People with strong agency tend to set goals and pursue them with confidence, even in the face of obstacles. They’re the type to say, “I’ll figure it out,” and then actually do it. On the flip side, someone low in agency might feel more like a passenger in their own life, waiting for external forces—like luck, other people, or circumstances—to dictate what happens next.

It’s not quite the same as assertiveness or ambition, though it can overlap. Agency is quieter, more internal—it’s the belief that you *can* act, paired with the will to follow through. Psychologists often tie it to concepts like locus of control: high-agency folks lean toward an internal locus, feeling they steer their fate, while low-agency folks might lean external, seeing life as something that happens *to* them.”

[Balkeum Labs @Balkeumlabs]

Thank you for having us 🙏
Balkeum Labs joined @DesciSeoul — great connecting with @paulkhls from @Bio_Protocol and @ZeusGlobalX from @AILiquid_

[Balkeum Labs @Balkeumlabs]

@Balkeumlabs joined the AI Native Enterprise Forum 2025 🚀
Big thanks to DLG & TrustConnector for hosting.
🔗 https://t.co/NW2L8WlULH

[DeSci Seou l: AI X Data @DesciSeoul]

Panel 1: DeSci × AI: How to leverage AI for Innovations and Research

Yijing Shi, CMO @GoKiteAI
Yoon Seokbin, Prof at Sogang University
David Mueller, Co- Founder @TheoriqAI
Francesco Andreoli, DevRel @MetaMask

[Balkeum Labs @Balkeumlabs]

https://t.co/4PU7FI139D

밝음랩스는 @OpenLedgerFdn Asia AI x Blockchain Grant Program의 첫 번째 수혜팀으로 선정되었습니다.

@OpenLedgerHQ는 단순 인프라 구축을 넘어 신뢰 가능하고 투명하며 검증 가능한 AI 경제를 위한 기반을 설립합니다. 온체인 Attribution 및 개인정보 중심 AI 스택은 강력하면서도 안전하고 사용자에게 소유권이 있는 헬스케어 AI 솔루션을 가능하게 합니다.

이 마일스톤과 함께 저희는 연합학습 기반 프라이버시 보호 헬스케어 AI 비전을 분산 AI의 가장 신뢰받는 리더와 함께 현실에 한 걸음 더 가깝게 만들고 있습니다.

[Balkeum Labs @Balkeumlabs]

🧠 What if every time your data was used… you got paid?

Not sold. Not exposed. Not exploited.

🔒 You keep your data.
🤖 AI learns from it.
💸 You earn.

Welcome to Balkeum Labs — where decentralized federated learning meets fair rewards.
Train AI. Keep Data. Earn Rewards.

[0G Labs @0G_labs]

Welcome to the sixth chapter of Guild on 0G Spotlight, a series highlighting projects that prove decentralised AI isn’t just theory.

This edition covers:
→ @Dormint_io: Decentralizing Wellness
→ @Balkeumlabs: Privacy-Preserving Federated AI
→ @HAiO_Official: AI-Powered Music for Web3

[Paul Kohlhaas bio/acc @paulkhls]

Did someone say DeSci summer?

[Eli Yune @eli_yune31608]

Privacy-preserving ML relies on a variety of PETs. Here's a breakdown of the major ones—how they work, when they're used, and what limits their deployment.

Differential Privacy (DP):
A mathematical framework that adds noise to training data or outputs to prevent leakage of individual records.
🟢 Pros: Strong privacy guarantees, even against inference attacks.
🔴 Con: Degrades model accuracy, especially in small datasets or sensitive tasks. Often used before training to sanitize data.

Secure Multi-Party Computation (sMPC):
Multiple parties jointly compute a function (like model aggregation) over their inputs, without revealing them to one another.
🟢 Pros: No raw data exchange; avoids reliance on trusted hardware.
🔴 Con: Communication overhead grows significantly with model size and number of participants.

Fully Homomorphic Encryption (FHE):
Enables computation directly on encrypted data. The result, once decrypted, is identical to computing on plaintext.
🟢 Pros: Maximum privacy; no need to decrypt during training or inference.
🔴 Con: Extremely slow. Hardware acceleration is still in early stages, making it impractical for most real-world workloads today.

Trusted Execution Environments (TEE):
Hardware-based isolated environments (like Intel SGX) that securely execute code and protect data in use.
🟢 Pros: Fast and easy to deploy on individual machines.
🔴 Con: Requires trust in hardware vendors; centralized by design. Vulnerable to side-channel attacks and lacks transparency.

Federated Learning (FL):
A protocol where clients train models locally and only share updates. Useful when data cannot be centralized.
🟢 Pros: Data never leaves local devices; useful in healthcare, finance, etc.
🔴 Con: Model updates are aggregated periodically, introducing delays due to waiting on stragglers and coordinating training cycles.

Each method contributes differently to privacy, utility, and decentralization. Often, hybrid approaches are needed—e.g., combining FL with sMPC (as in FLAI Protocol by @Balkeumlabs ) or adding DP noise on top of TEE inference. Design choices should reflect threat models, governance structure, and latency tolerance.

#FederatedLearning #DeAI #PETs

[Balkeum Labs @Balkeumlabs]

Big thanks to German Blockchain Week 🇩🇪 @gbweekofficial for having us on stage!
The @Balkeumlabs was proud to participate, share insights, and connect with builders shaping the future of Web3 + AI.

[Balkeum Labs @Balkeumlabs]

Great energy at the Blockspati Pitch Event during #GermanBlockchainWeek @gbweekofficial!
Proud to share what we’re building at @Balkeumlabs among visionary founders and tech leaders.
Huge thanks to The Block Spati and @Polkadot for hosting an incredible event!

[Chris Hladczuk @chrishlad]

Tattoo this to your brain:

"The world is a malleable place. If you know what you want, and you go for it with maximum energy and passion, the world will often reconfigure itself around you much more quickly and easily than you would think.”

– Marc Andreessen

[David Shapiro (L/0) @DaveShapi]

If you're a developer, these graphs should scare the ever living shit out of you.

Farming jobs: peaked in 1910 at 12 million jobs. Today? 2 million jobs. -80% jobs.

Farming output: +600% in the same time frame.

Manufacturing jobs: peaked in 1979 at 20 million. Today: 10 million. Half the jobs gone.

Manufacturing output: Doubled in that timeframe.

NO. JOB. IS. SAFE.

EVERYTHING. CAN. BE. AUTOMATED.

No, that doesn't mean it's "going away entirely forever" but what it *does* mean is that the total number of people needed to achieve the same output for any task can be halved, then halved again, and again.

When Microsoft and Google say that >30% of their code is now written by AI - THE. WRITING. IS. ON. THE. WALL.

[Andrew D. Huberman, Ph.D. @hubermanlab]

The health hierarchy:
SLEEP
EXERCISE
NUTRITION
STRESS CONTROL
HEALTHY RELATIONSHIPS
CIRCADIAN RHYTHM ADHERENCE
Bright mornings & days & dark nights = great sleep, energized focused days & overall robust mental & physical health.

[vas @vasuman]

I promise you you’re vibe coding wrong

as someone who has built multiple production-ready applications, with thousands of users, from just Cursor with minimum intervention.

But first here's you (probably):

You open Cursor. Type “build me X.”
It spirals. Nothing works. You start over.
That’s not development. That’s chaos.

I have an incredibly simple system that works every single time:

[Eli Yune @eli_yune31608]

This is just the start.

We haven’t even touched on what happens when FL meets incentives, coordination, and verifiability.

But we will.
Stay tuned.

[Eli Yune @eli_yune31608]

AI is hungry.
It learns from our voices, faces, locations, and health data.

But no one asked you.

There’s a growing movement to flip the script — and it starts with Privacy-Enhancing Technologies (PETs).

Let’s talk.

[Eli Yune @eli_yune31608]

If you’re building AI in 2025 and not using Federated Learning, ask yourself:
→ Why am I still centralizing?
→ Who owns the model?
→ What happens when users demand privacy?