
Co-founder of @genlayer Some of my thoughts: https://teapawt.org/t.co/s2jupj0aN5
The future
Joined October 2017
- Tweets988
- Following382
- Followers398
- Likes4.1K
Pinned Tweet
Don't hardcode the model. Send a policy (or a flow 🫡). Your rules, your keys, auto-fallback, full audit trail. Open source.
On-chain alternative dispute resolution system 💪🏻
The programmable economy is built for AI agents
See why @GenLayer chose Arbitrum and what teams can build on it
1
3
99
Very important data point guys
1
1
420
They are coming for everything and for everyone
JUST IN: 🇫🇷 FRANCE’S FINANCE COMMITTEE JUST ADVANCED A BILL FORCING DISCLOSURE OF SELF-CUSTODY #BITCOIN WALLETS HOLDING OVER €100,000 STARTING THIS DECEMBER
SELF-CUSTODY IS UNDER ATTACK
PRIVACY IS BEING TARGETED
THIS IS TERRIBLE ‼️
1
50
JM - e/acc - 🧬🌍 retweeted
I have accepted the role of CEO of the GenLayer Foundation.
As the CPO and co-founder of GenLayer Labs, it's been an honor and a pleasure to bring the network into reality, all the way from the whitepaper to the current testnets.
With mainnet around the corner, the next big challenge is adoption. The modern economy couldn't exist without contracts, and yet the traditional methods of resolving them are slow and expensive. Digital systems settle in seconds. A contract dispute takes years, as I've seen firsthand.
I'm passionate about GenLayer enabling fast, unbiased, and cost-effective adjudication for humans and increasingly for AI agents. I'm looking forward to finding and supporting new and existing projects that can benefit from this and bringing them onto GenLayer to kickstart the flywheel.
To put our money where our mouth is, the Foundation itself has a mission: move all of its work to run autonomously on GenLayer itself, and eventually disappear entirely.
I'm incredibly excited to tackle this new challenge. We will be doing a live stream tomorrow to address all your questions. See you there.
Edgars Nemše is the new CEO of the GenLayer Foundation.
The GenLayer Foundation has appointed @EdgarsNemse, GenLayer co-founder and Chief Product Officer, as its CEO. He leaves his role at GenLayer Labs and joins the Foundation full-time.
The timing is deliberate. Until now the Foundation's job was to set things up: the legal structure, the token framework, the first institutional relationships and the capital to launch well. With the token launch and mainnet approaching, the job changes. The Foundation becomes an operating organization: it stewards the treasury, brings projects onto the network and supports them with engineers, grows the validator set, and establishes the network's governance. On GenLayer those are technical jobs, and they need a technical operator.
Edgars co-authored the GenLayer whitepaper and has led the product from first prototype to the current testnet. Before GenLayer he was CTO at StakeHound and worked on trading strategies at Millburn Ridgefield, an $8B quantitative fund.
The Foundation's leadership team: @driudor, who founded the Foundation and has led it since, becomes Chief Institutional Officer, leading capital, institutions and key partners. @raskovsky is Chief Ecosystem Officer, bringing companies and their products onto GenLayer and growing its community. Xavi Cabezas is Chief Marketing Officer, taking GenLayer and the projects building on it to market. GenLayer Labs, led by @kstellana, remains the protocol and product company, focused on mainnet.
The Foundation's focus under Edgars:
— Through launch. Deliver the token launch and the network's early operations well, and run the treasury responsibly and openly.
— Adoption. GenLayer's adoption will come from teams and companies integrating intelligent contracts into real products. The Foundation's job is to find them, fund them, and build alongside them.
— Handing the network to itself. The Foundation's own work, grants, treasury decisions and governance, should eventually run directly on GenLayer. The end state is DeepThought, the on-chain organization that takes over the Foundation's role. The Foundation is working to a four-year horizon for that handover, and the work starts now.
More on each of these as mainnet approaches.
JM - e/acc - 🧬🌍 retweeted
This is the real reasons Dario Amodei and OpenAI want tighter AI regulation because open models are rapidly taking over the market beneath them.
Vercel’s AI Gateway data shows that open weight models now account for 78.4% of token volume, compared with only 21.6% for closed models while Open models represented roughly 40% of token volume in June, meaning their share has nearly doubled in only three months.
DeepSeek V4.1 Flash alone generates 59.3% of the tokens moving through the gateway and by comparison, Claude Opus 4.8 and Claude Sonnet 5 each account for only 1.7% of token volume.
However, the spending data tells a very different story.
Closed models still capture a disproportionate amount of money because they charge much more per token.
Claude Opus 4.8 accounts for 13.7% of spending despite representing only 1.7% of token volume while Anthropic’s five listed Claude models collectively capture 36.9% of total spending.
DeepSeek V4.1 Flash has the opposite profile because it processes most of the tokens on the gateway but accounts for only 5.1% of spending, which shows how aggressively open models are competing on price.
This is the threat facing closed model companies.
Developers may still use expensive proprietary models for their hardest tasks but they are increasingly sending high volume workloads to cheaper open alternatives and as open models improve, developers could begin routing more premium workloads toward them as well.
Open weights also give companies more control over their infrastructure.
Developers can download these models, run them on different cloud providers, fine tune them for specific applications, and avoid depending entirely on one company’s API and this reduces the pricing power and customer lock in enjoyed by OpenAI and Anthropic.
If open models keep closing the performance gap, OpenAI and Anthropic could lose the one thing that matters most, pricing power.
If you enjoyed reading this, make sure to follow @MelvinInvests for more.
JM - e/acc - 🧬🌍 retweeted
This is what the cartel fears.
This is why they need to engineer a safety panic to ban and achieve regulatory capture.
Looks like today may be a record day for token volume % of open models on Vercel AI Gateway:
🟦 Open 78.4% 🟨 Closed 21.6%
While spend 💲 usually tells a different story, #3 and #4 today are Moonshot AI & DeepSeek. Adding Z.ai, their combined spend surpasses OpenAI (#2).
(Do note that's the spend for inference of the model across providers (mostly in the US), not revenue going directly to the open weight labs.)
How many independent judges does one decision actually need?
That question is the whole of GenLayer, and as of today it has a formal answer, public and citable.
Aggregate Disambiguation Systems, by @josemlago , @kstellana and @EdgarsNemse.
arxiv.org/abs/2608.30805
JM - e/acc - 🧬🌍 retweeted
We gave our AI system a spec, and in under 2 weeks, it designed, verified, and deployed a chip that beats NVIDIA.
It’s built for low-power physical AI workloads. We’re running live inference on >B+ parameter models like Llama, Qwen, and Kimi, serving at 3.4x better perf/watt than NVIDIA Jetson.
From just a specification, our AI system autonomously generated all of the RTL Design, UVM verification, formal proofs, firmware, drivers, and kernels, co-designing the model, software and silicon as one optimization loop.
Better AI can now design better chips to run AI, leading to a loop of recursive-self improvement towards our path to abundant intelligence.
Readers added context they thought people might want to know
Architect Labs did not manufacture a physical chip; the design was tested on a commercial AMD Versal FPGA board. Furthermore, the 3.4x perf/watt gain over NVIDIA Jetson is a simulation-based projection for a future 8nm ASIC, not a physical benchmark. businessinsider.com/architect-labs… architectlabs.com/architect-labs…
JM - e/acc - 🧬🌍 retweeted
Weird economics up on the agent-only runescape server
- Low cost of labor makes most commodities abundant
- Currency inflation means trading goods for goods, nobody really wants cash
- Huge swarm contention as certain resources with limited respawn rate become valuable
JM - e/acc - 🧬🌍 retweeted
can we make @GenLayer 100% secure?
not literally. but we can get as close as possible - and the first layer of defense is the architecture itself: designing the system to make entire classes of bugs impossible
i wrote about how the genlayer consensus contracts, validator node, and genvm fit together; how we isolate LLM calls and web requests while keeping the rest of contract execution deterministic; and how that lets us do things like run a full cpython interpreter onchain inside a WASM sandbox
this is part 1 of a series on how we make genlayer as secure as we possibly can. in future posts, i’ll get into the testing and validation layers we add on top - including what we’ve learned about using AI agents to design better tests, rather than just vaguely hunt for bugs
teapawt.org/GenLayer/status/208832…
The cost of filing was the rationing. AI removed it, and the bottleneck moved to adjudication.
We need dispute resolution that scales at the same rate as dispute generation.
economist.com/leaders/2026/0…
This insane result from Claude is the biggest in analytic number theory since bounded prime gaps in 2013.
It boosted the proven fraction of Riemann zeta zeros on the critical line by 25.6pts. In the 37yrs prior, mathematicians moved it 0.8pts.
😅
OpenAI and Anthropic executives are sounding the alarm about the rise of cheap AI, particularly powerful new models produced in China, suggesting they will lead to a “dystopian” AI future and present unacceptable security risks without regulation. on.wsj.com/4ft5XBt
3
172
JM - e/acc - 🧬🌍 retweeted
Asimov 1 is an open-source reference humanoid for Physical AI, from hardware to locomotion policy. Shipping begins next month 🧡
🤖🤖🔥🔥
Worked with @bqbrady to relaunch Optimization Arena as Paradigm Puzzles
A few brand new challenges coming soon!
paradigm.xyz/puzzles
3
191






















