Kernel (@Kernel0x) header
Kernel (@Kernel0x) profile picture

Technology as craft. Entering KB12, our 12th block. Applications always open: https://nitter.cf/t.co/ymO1dwuta4

Joined October 2020
Photo by Kernel (@Kernel0x)Photo by Kernel (@Kernel0x)Photo by Kernel (@Kernel0x)Photo by Kernel (@Kernel0x)Photo by Kernel (@Kernel0x)Photo by Kernel (@Kernel0x)Photo by Kernel (@Kernel0x)Photo by Kernel (@Kernel0x)Photo by Kernel (@Kernel0x)Photo by Kernel (@Kernel0x)
Kernel retweeted
🫡 Honcho 3.0.0 is here. Continual learning for stateful agents. Give your agent the 10K tokens that matter, not the 100K that don't. Give it memory that reasons. ⚡ SOTA & pareto dominant ⚡ 5x cost reduction ⚡ Per-token pricing ⚡ Unlimited retrieval ⚡ 60-90% less tokens ⚡ Granular reasoning control Better context. Fewer tokens. Any stack. Any scale. Much more 👇
2
8
5
29
7,523
Kernel retweeted
this is a really big release for us 🚀 we've productionized the version of honcho that defined + dominates the pareto frontier in agent memory biggest things to note: --docs overhaul. if you've felt confused by honcho in the past, pls give it another shot. if it's still confusing, lmk --we've been saying memory is reasoning, so now it's priced like a reasoning model. $2/m input tokens --variable output pricing. reasoning efforts now available --our landing page got a massive revamp that explains all of this too??!! check out honcho [dot] dev
🫡 Honcho 3.0.0 is here. Continual learning for stateful agents. Give your agent the 10K tokens that matter, not the 100K that don't. Give it memory that reasons. ⚡ SOTA & pareto dominant ⚡ 5x cost reduction ⚡ Per-token pricing ⚡ Unlimited retrieval ⚡ 60-90% less tokens ⚡ Granular reasoning control Better context. Fewer tokens. Any stack. Any scale. Much more 👇
3
4
24
1,880
Kernel retweeted
In 1993, Wired launched with Marshall McLuhan on their masthead. Silicon Valley loved “the medium is the message.” McLuhan was famously faith-driven, a converted Catholic. He (alongside Ivan Illich) have served inspiration for @left_pad. Best known as a steward of the pivotal open-source JavaScript compiler Babel, Henry (KB4) has spent much of his life creating open source tools – a social annotation layer for the web, a receipt printer, a faith-inflected OSS podcast. Throughout, he remains keenly aware of the effects of his “electric engineering.” “Electric information environments being utterly ethereal fosters the illusion of the world as spiritual substance,” McLuhan wrote. “It is now a reasonable facsimile of the mystical body, a blatant manifestation of the Anti-Christ. After all, the Prince of this World is a very great electric engineer.” Those searing words anchor @left_pad’s (KB4) nascent Bible App. “When God creates everything, he says that he creates humans in his image,” Henry told Kernel. “What that means is that we are meant to create.” His feed-based Bible App is a new message in a familiar medium. It’s a tool built to subvert TikTok “brain rot,” and to create precedent for technology that provides more signal and less noise. Cursory McCluhan readers may be surprised that he had more to say on his most known quote. "In Jesus Christ, there is no distance or separation between the medium and the message,” he wrote in the 1974 essay, Liturgy and the Microphone. “It is the one case where we can say that the medium and the message are fully one and the same." Henry tells his story of where the medium & message intertwine, and along the way, encourages each of us to create our own.
5
9
35
4,747
Introducing "Fast-fourier Transform (FTT), Number-theoretic Transform (NTT) and Lattice-based Cryptography", a two-part series exploring FFT, NTT and how NTT is used in lattice-based algorithms like ML-KEM and ML-DSA. godspowereze.com/posts/fast-… I'm happy to share the part one of this focused on FFT and NTT. With over 7500 words and rich diagrams, it takes you on a journey from Polynomial Multiplication, Complex Numbers and Roots of Unity to FFT and NTT. It makes no assumption about your knowledge. Rather, it gives you everything you need to follow through. All you need is the willingness to learn. I put in so much love and care into making this valuable to you and I hope you get value out of it. Do check it out and let me know.
2
7
31
1,074
Kernel retweeted
Replying to @Kernel0x
Glad this popped on my feed applied!
1
2
300
Thrilled to share that I've joined the @ethereumfndn as its in-house defi specialist. I’ll be working to deepen the Foundation’s connection to defi and strengthen Ethereum as the global financial settlement layer. Onward 🫡
60
25
1
324
45,412
The leader in Agent Memory. Top scores on 3 leading benchmarks + the most token efficient memory on the market. Always ahead of the curve. gg @plasticlabs crew.
Honcho has fully determined the Pareto Frontier in Agent Memory. It’s state-of-the-art on LoCoMo, LongMem, and BEAM. Honcho is also the fastest, cheapest and most token efficient memory solution.
10
9
1
32
2,354
Honcho has fully determined the Pareto Frontier in Agent Memory. It’s state-of-the-art on LoCoMo, LongMem, and BEAM. Honcho is also the fastest, cheapest and most token efficient memory solution.
32
47
22
262
66,607
Kernel is a great program and community. Applications are open for their next batch!!
Kernel is a residency for technologists pursuing high-quality work – from research to production – in crypto & beyond. We offer: ▷ A technical & philosophical culture ▷ Accountability, refinement, & distribution ▷ A network of unusually talented peers KB12 apps are open. Apply early by 12/31. apply.kernel.community/
2
1
5
1,498
Kernel is an incredible place to build — if anyone has questions or wants an intro, please let me know 💖
Kernel is a residency for technologists pursuing high-quality work – from research to production – in crypto & beyond. We offer: ▷ A technical & philosophical culture ▷ Accountability, refinement, & distribution ▷ A network of unusually talented peers KB12 apps are open. Apply early by 12/31. apply.kernel.community/
1
3
271
the 12th time is the charm looking forward to working closer than ever with the participants in @Kernel0x the energy shift from fellowship to residency is palpable if you're considering 'what is next?' or building something new, we'd love to chat nitter.cf/Kernel0x/status/200205…
Kernel is a residency for technologists pursuing high-quality work – from research to production – in crypto & beyond. We offer: ▷ A technical & philosophical culture ▷ Accountability, refinement, & distribution ▷ A network of unusually talented peers KB12 apps are open. Apply early by 12/31. apply.kernel.community/
2
11
496
Replying to @Kernel0x
@Kernel0x's back in 2026 for a new -- yet familiar -- residency program as per usual, it's grounded in craft, quality, and a sense of a more intentional, collective future
Kernel is a residency for technologists pursuing high-quality work – from research to production – in crypto & beyond. We offer: ▷ A technical & philosophical culture ▷ Accountability, refinement, & distribution ▷ A network of unusually talented peers KB12 apps are open. Apply early by 12/31. apply.kernel.community/
1
3
255
As a #KB5 fellow, strongly recommend Kernel to any tech builder curious about web3/crypto & beyond! It'll expand your thinking, spark new ideas, and connect you with an incredible community of like-minded folks in profound ways.
Kernel is a residency for technologists pursuing high-quality work – from research to production – in crypto & beyond. We offer: ▷ A technical & philosophical culture ▷ Accountability, refinement, & distribution ▷ A network of unusually talented peers KB12 apps are open. Apply early by 12/31. apply.kernel.community/
1
2
290
Today we’re announcing a $29M Series A extension for LI.F​I, funded by Multicoin and CoinFund. This brings our total capital raised to ~$52M. More aggressive and committed than ever to build the universal liquidity market for digital assets. Time to scale. fortune.com/2025/12/11/exclu…
189
43
48
658
268,484
⚠️Get a free month of Honcho Chat! ⚠️ First 50 to rt or comment get a code. Build identities. Build themes. Build artifacts. Build your representation. See how your preferences shape them. Experience AI memory infrastructure that actually remembers you. Upload your Chatgpt history or just start chatting Happy Cyber Monday!
11
9
2
21
1,933
I’ve had The Beginning of Infinity unread on my shelf for a while. It feels like a good explanation. It reminds me of @Kernel0x’s education philosophy – peer learning, “unschooling” society, by recognizing that appealing to authority for truth limits us all … Time to read …
The most radical idea I learned from The Beginning of Infinity: All human progress comes from one thing. Not talent. Not experience. Not gurus. The creation of good explanations. Once you see this, everything changes. @DavidDeutschOxf 0/15
1
3
584
Listen to this if you want to know more about pod! The description on the tweet is also a good TLDR
What if we don't *need* global consensus to solve the double spend problem? ▷ validators don't talk to each other directly, instead maintaining a local, totally ordered log ▷ transactions finalize in one round-trip (200ms), without being ordered ▷ ordering-sensitive applications can build custom sequencing gadgets, on top In this log-based structure, consensus is deprioritized against finality. We move from 'blockchains' to 'blocklessness', moving the cycles onwards. The @poddotnetwork team (led by @sagrawal @abresas) has been exploring this design space and other 'golden geese' in crypto (like sparse clients), that often go understudied. They are worth the follow if you're looking for builders with strong opinions and understanding of the crypto design space.
2
2
9
1,032
s/o to @Kernel0x for hosting me for this talk focused more on how reasoning about identity could become a protocol ultimately powering a user-owned data layer tons of fundamental ML and product challenges to solve before then, but we're closer than we think
The context wars are upon us. OpenAI, Anthropic, Apple, Google. Every major player is racing to collect more of your data to power their AI. It reeks of “enshittification.” At Kernel's salon in NYC, @plasticlabs CEO @vintrotweets laid out a different vision: Honcho is open-source memory infrastructure where your virtual identity stays yours, while agents benefit from state of the art long-term memory. The thesis: memory should be a commodity, not a moat. Give it away for free. Instead of treating memory as storage/retrieval, their Neuromancer reasoning models predict and deduce per-user context, training on bespoke outputs with rich personal context that foundation models can't access. Honcho has 620k+ messages processed since soft launch 2 months ago. Want to help prove this works? Grab a free API key: app.honcho.dev
2
2
6
616
The context wars are upon us. OpenAI, Anthropic, Apple, Google. Every major player is racing to collect more of your data to power their AI. It reeks of “enshittification.” At Kernel's salon in NYC, @plasticlabs CEO @vintrotweets laid out a different vision: Honcho is open-source memory infrastructure where your virtual identity stays yours, while agents benefit from state of the art long-term memory. The thesis: memory should be a commodity, not a moat. Give it away for free. Instead of treating memory as storage/retrieval, their Neuromancer reasoning models predict and deduce per-user context, training on bespoke outputs with rich personal context that foundation models can't access. Honcho has 620k+ messages processed since soft launch 2 months ago. Want to help prove this works? Grab a free API key: app.honcho.dev
2
5
1
17
1,929