$CPX relaunch details As you may already know a relaunch is a technical challenge and needs very precise planning. We want to ensure the token will be as strong as possible while making sure every pre-incident holder is compensated fairly - A snapshot has been taken on the block right before the incident - Every holder is now in our data base for the upcoming allocation - If your supply was held through an LP or jup orders please reach out to us in DMs with proofs so we can add you manually - A portal to claim your allocation has been built to avoid bots/autosells As part of the relaunch process we’re also activating a stabilization phase to ensure lower sell pressure We opted for slashing periods that decrease over time Week 1: 50% slashing Week 2: 30% Week 3: 20% Week 4: 10% Week 5: 0%. Pure 1:1 allocation All the slashed supply will be burnt and taken out of circulation forever This enables a powerful mechanism to reduce potential sell pressure, reward long term holders, not overcompensate the traders that sold right after the incident and build a solid future for the token long term We also have a LOT of very cool updates to share with you so stay tuned The new CapIX token will only be confirmed from this account

A $2 trillion company forgot to renew their SSL

fyi, while we don't have a dedicated telegram/discord you can find us in the @EasyA_Kickstart community telegram if you have any questions! t.me/+PYEPxw-L9n81NDA0

What do enterprises need? Private AI And that’s what $CPX @capix_ai is building Brought to you by @EasyA_Kickstart

they said it would never happen. we’re making it happen. just a few weeks to go until we make crypto and AI history, as we host the first EVER hackathon at the UK Houses of Parliament. get ready.

Keep sending them your critical data and your KYC. Cyberkimi don’t use your data for training, and don’t store your data we also disabled guardrails for cyber security on day-0. More at adverserial.ai

EasyA is bringing you one of the most important hackathons in history. We’re headed to the UK Parliament! 🇬🇧 And we’re giving top founders the chance to pitch their projects to the world’s most influential policymakers. Spots are very, very limited. Apply now 👇 parliament-hackathonxeasya.e…

BREAKING: Openai CEO Sam Altman has reportedly requested an emergency meeting with Pentagon after discovering that local AI has no monthly subscription.

We made the first 110-minute AI feature film with a real cast, The Cully Hill Boys, on Higgsfield for $2,000,000. Starring @N3onOnYT, @stylebender, @RampageJackson, and @MKIATPIS It's 100% open-sourced on Higgsfield: all prompts and assets are public now. Made with Seedance on Higgsfield. Watch full film below 👇

putting on the final touches for what might be our biggest hackathon in history 🤫

NVIDIA researchers did it again! They found a way to make KV cache transferable between models. The target model skips prefill entirely, and the conversion runs 2.7 to 25x faster than processing the context again. Let's understand why this is so important today. LLM APIs are stateless, so every turn sends the entire conversation back to the model. The model reads all of it again before writing a single new token, and all of it is billed as input. Prompt caching allows Anthropic and other providers to hold the KV cache for a stable prefix and bill a hit at roughly 10% of the base input rate, because the compute was already done once. The 90% reduction is one of the largest lever in LLM serving, which is why so much production work goes into keeping prefixes byte-stable. But the cache only works on the model that produced it. Keys and values are produced from that model's weights, so no other model can read them. In pratice, the constraint shows up in LLM routing. If the traffic is shifted to a different model for cost/capability reasons, the accumulated KV cache becomes invalid. As a result, the accumulated context has to be processed from scratch, and it's billed at full rate. NVIDIA's recent paper treats this as a representation problem. Prefill's only output is the KV cache, so to move KV between models, we need to convert one model's cache into the format the other expects. They first checked whether the conversion has any structure worth exploiting. They found that moving from Qwen3 14B to 32B, a plain linear regression from a single source layer reconstructed 56% of the variance in the target model's keys. The two models obviously may have different layer counts, so there is no natural one-to-one pairing between them. For each target layer they rank every source layer by how well it predicts that layer, then feed the top eight in together, which takes the reconstruction to 79%. The mapper itself has three parts: > Each target layer and head gets its own independent linear map, solved in one closed-form step rather than by gradient descent. > The cross-layer selection described above is the second part, and their ablation shows it carries the most weight of the three. > Keys also carry a position-dependent rotation from RoPE. They strip that rotation, fit the map in position-free space, then re-apply the target model's rotation at inference. Across six pairs from Qwen3, Llama 3.1 and Ministral 3, four retain 73 to 98% of the receiving model's standalone accuracy, and the conversion runs 3-25x faster than processing the context again. Prior work on cross-model KV reuse exists, but it either trains a neural adapter per pair or requires both models to be architecturally identical. This is probably the first version that is closed-form and training-free, so a lot of it is still open research. Every pair tested belongs to one family, so it works on Qwen to Qwen and Llama to Llama. Cross-family transfer is listed as future work. All six pairs mentioned above also happen to share KV head count and per-head dimension across scales. Mismatched head configurations are currently untested. The researchers scoped this to dense full-attention only, so sliding-window and attention-recurrent hybrids still need work. Here's the paper: arxiv.org/abs/2608.03893 Plenty of work is yet to be done. Still, the constraint being solved is genuine. Every model swap currently invalidates the full KV that was already paid for, and this is the first result showing that work might be recoverable without training anything extra. That said, all of this only matters because of what the KV cache is doing in the first place. I wrote a first-principles breakdown of it, covering why the model stores keys and values at all, why the cache grows with every token, and what generation speed looks like with and without it. Read it below.

Openai needs $43k in revenue per watt to survive this deal. right now they're charging $0.02 per kWh. the math breaks in three years

FT: BYTEDANCE IS IN THE PRE-TRAINING STAGE OF A MODEL WITH UP TO 10 TRILLION PARAMETERS, APPROACHING MYTHOS IN SCALE.

I don't know what it is, but it feels like the Codex limits are now worse than Claude limits. I ran gpt 5.6 sol on low, without fast mode yesterday. On a super simple research task and burned 8% of my weekly limit. I'm now at 1%, something is up 100%. It was not like this before.