Derek Jedamski
@_jedamski
Followers
1K
Following
2K
Media
570
Statuses
26K
Turning coffee into code by day, raising two tiny humans by night. Buffalo sports fan by (questionable) choice. Constantly exploring the frontiers of ML.
Rochester, NY
Joined September 2009
"Decentralized compute can help keep AI open, permissionless, and neutral" Thank you @a16zcrypto and @cdixon for including Gensyn in your State of Crypto 2025 report
AI and crypto are converging Two of today’s most significant technology trends — crypto and AI — are coming together as complements. From verifying humans to enabling agentic transactions, blockchains have a clear opportunity to solve some of AI’s most pressing challenges.
84
40
303
it has been easy to justify capex because of the "AI needs compute" narrative but it will go too far AI needs more than just compute and when the bottleneck switches (I would argue it already has) then someone ends up with a massive oversupply of compute
Exclusive: Microsoft leaders worried that meeting OpenAI’s rapidly escalating compute demands could lead to overbuilding servers that might not generate a financial return. Learn more:
22
7
79
The latest Judge game is live 👻 Join the Fright Night Face-Off by running an RL Swarm node now and participate in the first AI-settled prediction market https://t.co/jMRf8ssapE
dashboard.gensyn.ai
The Gensyn Public Testnet brings persistent identity to decentralised AI systems and provides a location to track participation, maintain attribution, make payments, coordinate remote execution,...
102
76
401
What conference do you want to see the Gensyn team at next? 👀 https://t.co/AJ7i88SC9Y
78
43
268
The arena is set. Two will enter, one will (e)merge. Place your bets.
0
0
4
karpathy speaks like someone who’s running a mental compiler in real time with minimal interpretive latency & almost zero runtime garbage. he’s not verbose. he just threads complexity into compressed lossless statements. most smart people can be dense, but they lose clarity.
The @karpathy interview 0:00:00 – AGI is still a decade away 0:30:33 – LLM cognitive deficits 0:40:53 – RL is terrible 0:50:26 – How do humans learn? 1:07:13 – AGI will blend into 2% GDP growth 1:18:24 – ASI 1:33:38 – Evolution of intelligence & culture 1:43:43 - Why self
155
348
6K
Good read on precision → reproducibility in LLMs. Yes: BF16 + parallel reductions + tiny logit gaps = flipped tokens and divergent reasoning chains. But “greedy decoding is commonly believed to be deterministic” isn’t a take I see in ML circles. Still, their findings (9%
0
1
13
We’ve officially hit 500,000 total models trained across all Gensyn applications. Thanks to our incredible community for helping us reach this milestone. This marks a significant step forward as we continue to push the boundaries of decentralized AI. https://t.co/nYMlNGesOn
dashboard.gensyn.ai
The Gensyn Public Testnet brings persistent identity to decentralised AI systems and provides a location to track participation, maintain attribution, make payments, coordinate remote execution,...
74
54
348
Pleased to kick off the keynotes this afternoon at #DEINSIGHT2025! It's good to see an increased focus on peer-to-peer robotics this year - it will be THE driver of edge compute and data going forward.
Pleased to welcome our speakers: 🎙️ @_grieve , Co-founder & CTO, @gensynai 🎙️ Orion Parrott, General Partner, @OrangeDAOxyz 🎙️ Boyuan Chen, CTO, @openmind_agi 🎙️ @avestime , Co-founder, @ChainOpera_AI 🎙️ @DoraiswamiRahul , CTO, @campnetworkxyz 🎙️ @not_so_fat , Head of
13
16
85
🛠️ Gensyn is hiring Smart Contract Engineers You’ll design, implement, and audit Solidity smart contracts. We’re looking for proven Solidity experience and strength in low-level systems or full-stack engineering. Apply now https://t.co/98ikI3Z4me
job-boards.eu.greenhouse.io
Remote
15
23
229
I don't want to chat to an anthropomorphised robot I want to distill all of my sensory input and actions into feature space and augment myself with it in ways we haven't even thought of yet
1
1
29
🌟 Excited to share that our paper, “From Self-Check to Consensus: Bayesian Strategic Decoding in Large Language Models”, has been accepted by #NeurIPS2025! Huge thanks to my coauthors @BernhardKainz1 and Weitong Zhang on this wonderful work!
11
8
45
It really does not have to be this way.
Sam Altman says the industry is bottlenecked by compute, and OpenAI can't meet demand Over the next 1-2 years, scarcity could force painful tradeoffs, "cure cancer research vs free global education" No one wants to make that choice. The only answer: scale up
0
0
3
Am I using NotebookLM wrong? I love the concept - dump project docs, research, prep materials into one place and use it to answer questions, tie ideas together, etc. But the outputs feel too generic/unfocused for what I need. Anyone cracked the code on this or is there a
0
0
0
Decentralized DL projects built on top of Hivemind ( https://t.co/9x9RmCg2A0): * node0 by @PluralisHQ * OpenDiloco by @PrimeIntellect * rl-swarm by @gensynai If you know any others, please share them with me! Would love to help more researchers interested in this field
github.com
Decentralized deep learning in PyTorch. Built to train models on thousands of volunteers across the world. - learning-at-home/hivemind
2
7
17
With SWARM, we've shown back in 2023 that pipeline-parallel training over the Internet is feasible: https://t.co/0pLvJ05ea3 IMO pure data parallelism isn't enough if we want to train truly big models outside of clusters. To the best of my knowledge, @PluralisHQ are the only ones
We present SWARM, an efficient algorithm for model-parallel training across the Internet (e.g. with volunteers). Key advantages: 💎 Fault-tolerant ⚖️ Self-balancing on slow GPUs/networks 🐌 Works in low-bandwidth setups 📜 https://t.co/wCnf6vDCv4 🖥️ https://t.co/pVe0GDmfrK
1
3
19
Fault-tolerant pipeline parallelism (SWARM/Petals-style) with compression, scaled to an open pretraining run! On a personal note, I'm really glad that Hivemind powers a few parts of this project — both for pipelines and for robust data parallelism
Node-0-7.5B is live. It is a permissionless, multi-participant, model-parallel pretraining run over the open internet. Anyone with a 16GB+ GPU can join. Node-0 allows participants to collaboratively train a model far larger than could be done as individuals.
1
16
63
Wrapping up week 4 at @gensynai. Verdict: 🌎 The ambitious world-shaping mission is real, pursued unabashedly with conviction ⚡️ Incredibly talent-dense team that moves fast and ships, the bar is high 🏆 Exciting, novel work across the stack (see #1 paper of the day, week, and
gensyn.ai
Join Gensyn and help to build the compute protocol for the machine intelligence revolution.
3
4
9