peter kris retweeted
X intro: I'm Matt, a Czech-Slovak guy in red glasses, programming plant DNA to build the first Crop Design Company, to make GMOs sexy again, eventually :)
Funny enough, my family were in Greenpeace, the biggest GMO haters. I wanted to be a conservationist and joined some NGOs, but also crypto-anarchy groups, where I realized the best way to protect nature is to own land privately and leave it be.
Biotech was my closest shot at money, since I won some national biology competitions. So I went into cancer research, where the money is, until I met a girl, a plant scientist. Didn't get the girl, but became the plant guy who led a student team winning iGEM, the Olympics of biotech, together with @matgries and @miro_slau, Nitroduck's co-founders :)
Before Nitroduck took over, I held Slovakia's birding Big Year record, was the country's most productive bike food courier, a landscape photographer, mountaineer, and crypto anarchist :))
Now I'm in the US, forever grateful for @ycombinator , 5050, Tesla FSD, and so much more this country gave me!
the age of ideological LLM models is upon us
Introducing Arete-27B: Qwen3.8-27B, post-trained to reinforce Western values.
Chinese open models are currently the strongest available, but they ship with state-aligned behavior in the weights. We built an 803-prompt eval across seven value categories, including freedom of expression, democracy, and privacy, and measured how often a model adopts China’s party line.
After fine-tuning on roughly 53 million tokens, Arete-27B's propaganda rate fell from 22% to 1%, in line with the Western models we tested. Its Hermes Bench score shows no significant performance tradeoffs.
The values in an open model are one fine-tune away.
Our eval, weights, and training data are available now along with the full blog post: nousresearch.com/arete
4
183
peter kris retweeted
Deriverse just became the largest USDT/USDC venue on @solana.
$51M in 24h. More than all the CLMMs combined.
High performance onchain orderbooks are finally here 🪐
Deriverse is now the largest USDT/USDC venue on @solana.
$51M in the last 24 hours.
We're just getting started 🪐
ALT Source: solscan.io
pre-seed and seed is still relevant everywhere. If you're a VC and playing a seriesA+ game, its more resembling politics and that is concentrated in SF. Just like DC
Sequoia just said preseed/seed isn't important to them or that the real top game was never outside SF
Sequoia saying what those in the Bay Area have known for the last 5 years.
The only relevant cities for venture today are SF and TLV.
1
2
222
Ethereum should be at least 10x asset from here.
Aave as lead lending protocol should easily be 100x
Decentralized protocols are global foundations for finance and they shouldn't be worth just as another company
just another company is worth 500B in 2026
Make it go higher
1
15
529
peter kris retweeted
🚨 Huge news 🚨 today we are launching SIE Serverless - a managed inference product focused on small models. It’s useful for spiky workloads on 20+ task specific workloads across Search, OCR and Agentic execution.
Sign up for a $500 credits launch grant 👇
It’s powered by our open source inference engine that has received thousands of github stars in the last few months, battle-tested across finance, legal & tech. We believe that self-hosting is going to be critical given all the trust & security issues in AI, but practically there are always workloads where you’d like somebody else to deal with the GPU quota like synthetic data generation, evals, public research and beyond.
The team at Superlinked has been burning the midnight oil for months and I’m so proud of every single one of you - what a privilege to call you my team mates y’all!
We built it, now help us make some noise 🙌
A like, a comment or a repost would mean a lot! ❤️
Median price of real estate in Prague downtown per square meter is appr. 11k USD
Thats very close to median San Francisco
It means only two things - either Prague is extremely overpriced or San Francisco is underpriced
Pick your macro pill
5
8
311
It's quite a revolution seeing how AI can 100x the new plant breeding
They modify the pollen with CRISPR and then pollinate the crop. Just think of it like new super-powered cocoa or olives that is naturally resistant to fungi.
The best thing is it's much more natural than using chemicals (pesticides) and it solves the current rampant issue of crop production decline globally
Nitroduck designs new colors, flavors, and disease resistance into fruits and veggies. We integrate CRISPR into breeding through pollen, making any crop gene-editable.
Breeding a new specialty crop variety takes 8 - 16 years: breeders hope for random gene combinations and mutations, then spend years picking out the good ones. Meanwhile, consumer tastes, the climate trends, and pathogens change way faster.
Now, accessible genome sequencing and AI can tell breeders what to aim for instead of what to hope for, and CRISPR can turn random mutations into chosen ones. The catch is that CRISPR only works on about 15% of crops. It has to go into a single stem cell that’s then regrown into a whole plant, a process that has to be worked out for each variety, which is cost-prohibitive.
We skip the stem cell and put CRISPR into a pollen grain. The grain pollinates a flower, and out comes an edited seed - seamlessly tapping into how breeders already work.
Matúš, Miro, and I are plant synthetic biologists who made duckweed, the smallest and fastest-growing plant on Earth, genetically engineerable. For doing so we won the Grand Prize at iGEM, the “Olympics of biotech.” During our studies, we engineered 4x more plants than the average plant scientist does in a whole career. Now we’re building Nitroduck, The Crop Design Company, which will make pesticides obsolete by 2042.
4
8
404
Mind you, the normal more technical people are only now buying mac minis. It's been 8 months after openclaw craze when X was all about macminis!
If you think context and company brain is the king now, you are way way ahead.
1
3
180
AI will have much less impact in Bio
the point is not only the quantity of stuff and complexity in biology. The main thing is biology actively changes in response to any action you or AI takes.
It changes in a way that breaks your original fundamental assumptions, so the actor is forced to go back to drawing board
its more game theoretic problem and less knowledge / understanding / efficiency problem
1
2
192
I just had an aha moment why @daytonaio is useful
was scratching my head for a while but now it all makes sense
The most annoying issue with current agentic models is their hygiene.
The way they pollute the system environment, don't rotate logs, archives and dumps, install packages uncontrollably, etc., really shows how they were trained in throwaway VMs and it really affects usability.
2
4
211
peter kris retweeted
The most annoying issue with current agentic models is their hygiene.
The way they pollute the system environment, don't rotate logs, archives and dumps, install packages uncontrollably, etc., really shows how they were trained in throwaway VMs and it really affects usability.
Internet exploits will be a false alarm just like exploits in crypto were 6-7 months ago
Crypto was at peak FUD, exploits every week, and then suddenly it all calmed down. AI took the low hanging fruit coming from new models, and the same is happening now in broader internet
2
163
peter kris retweeted
If you think this, then you haven’t worked AI in a niche legal problem. AI largely fails where there isn’t a lot of prior data. In the patent world (where shit is new almost 100% of the time) or in other novel legal contexts, AI WILL struggle.
Might be surprising to someone, but law firms adopted AI very soon and got familiar deeply, because they had the first writing on the wall
Heard of a top law firm completely cutting off Harvey and all API calls to Claude/OpenAI after they saw what openAI did with navier strokes
Now hosting everything locally with GPUs they own, running Kimi and deepseek
Think you’ll see a lot more of this going forward
2
166
I was invited to an exclusive private event where 21 CEOs and CIOs from traditional industries talked about their AI implementations.
My biggest takeaway: AI enterprise adoption is in much better shape than I expected.
There were people running manufacturing plants, logistics, agricultural businesses, pharma and industrial services.
Here’s what stood out:
1. Every company has internal AI lead that’s focusing on very practical automation
Processing orders, scanning invoices, double-checking POs and contracts, turning messy emails into structured data.
2. Some teams are vibe-coding their own domain-specific harnesses (good luck or rip latest YC batch)
I’ve seen apps with custom chats, pluggable frontier models and open-weight models, custom desktop apps enhanced through small AI side-chat. Internal people know the practice very well, so they already have ideas of what should be running in production and which workflow to automate.
External FDE companies with domain-specific workflow knowledge don’t stand a chance against industry veterans.
3. Unverified AI outputs with questionable quality
Different references for the same shipment. Missing invoice attributes. Inconsistent documents. Historical prices without context. Hallucinations are still a thing when challenged by the rigid practical experience.
Some eval framework or guardrails are missing.
4. “We saved time” needs a second sentence.
Did you reduce an external bill?
Avoid a planned hire?
Handle more business with the same team?
Improve turnaround time?
Increase margin?
Several presentations reported dramatic improvements in speed. The connection to financial results was often much less developed.
5. Physical automation is its own universe
Production volume, changeovers, reliability, maintenance and integration can completely change the return on the same category of robot.
The candid stories about failed automation were among the most useful parts of the day.
Collaborative robots are important, so robots can co-exist in the same physical space as humans without the risk of physical harm.
6. Human review is still doing essential work.
People openly described incorrect outputs, failed experiments and cases where someone trusted an answer too quickly.
7. Everyone is building their company’s context / knowledge base
Teams are recognizing the value of context and having it pluggable to any model in the future. Reduce the dependency on single model and have the flexibility to switch
8. Little trust to frontiers
Everyone acknowledges it’s problematic to trust frontiers with their data. ZDR policies are in place, but technically it’s hard to trust it anyway. It grows a need for own cloud, or separate cloud on hyperscaler’s instance
9. Growing recognition of need for process formalization
They know their workflows and have ideas for automation. But they begin to realize that to do it in a systematic way, you need to map the processes in the company first, look at them on high-level and decide for restructuring and then automating where it fits the best.
All in all, the state of adoption is in good shape, and this makes me more bullish on AI growth and upcoming Anthropic IPO.
3
3
225
peter kris retweeted
we could treat AI as just humans that can do stuff digitally. We could return mentally to like year 1850, where everything was like work with humans and do stuff for other humans
In that we maybe shouldn't try to automatize guardrails, just hire better human managers of agents
peter kris retweeted
stupid thing I noticed, I sometimes speak/write not X, its Y, just naturally without Claude help. More than I used to
I observe how my brain fry in real time
I expect more countries to adopt EURo
Bonds are rising % yield, so public debt is getting more expensive. If your country adopts EUR, you suddenly have lower more stable ECB rate and in effect your public debt gets lower
Hungary has 5.50% central bank rate, and if they get euro, it will fall to 2.65% and in effect their public bonds 5.357% will get repriced lower too
