Register and share your invite link to earn from video plays and referrals.

Search results for byteS
byteS community
One keyword maps to one global community path.
Create community
People
Not Found
Tweets including byteS
making search fast means touching fewer bytes and doing less work per byte tpuf v3 compresses columns with composable schemes: 1M zeros becomes {val: 0, len: 1_000_000}, and predicates run directly on compressed data 1 comparison vs 1M → faster filters
Show more
Post-quantum transactions that require only 96 bytes on-chain: @will_wjk (Imperial College) et al. published a very interesting variant of Shielded CSV, called Compact Shielded CSV. I've always thought the CSV model is attractive in a PQ setting: the blockchain only stores the nullifier, so miners don't have to verify a signature before it gets into a block. Compact Shielded CSV goes a step further: the on-chain nullifier carries no signature at all, only what they call a "degriefer". Pre-quantum that's more than Shielded CSV's 64 bytes, but post-quantum it's much less. Unlike Shielded CSV, which operates on top of Bitcoin, Compact is its own blockchain. That matters for the degriefer: a Compact miner who includes a nullifier with an invalid degriefer can't claim the block reward. Shielded CSV could adopt the degriefer, but on Bitcoin there is no such penalty. Anyone who sees your nullifier before it's confirmed can front-run it with bogus copies for ~100 bytes of fees each, and your spending proof then has to show for every one of them that it's invalid. This is something we deliberately avoided in the design of Shielded CSV. With PQ signatures as large as they are, it probably is a reasonable trade-off. Compact also never discusses reorgs, which are a significant part of the Shielded CSV design. Integrating the degriefer into Shielded CSV would require combining it with the existing reorg handling.
Show more
new: i8 vectors f32: 4 bytes/dim i8: 1 byte/dim 4x fewer bytes → 75% lower storage and query costs + faster queries when embedded with a quantization-aware model (e.g. voyage-4-large) trained on i8 vectors, recall loss can be ~0! docs:
Show more
In the shadows of bits and bytes, a secret alphabet blooms - petals unfolding to reveal hidden truths
Nature research paper: Retrofitting language models to operate over bytes
I want to see DeepSeek's long context evals 890 bytes per token is insanely low
tripleS Event Gravity tripleS Character Name 🏅Final Result byteS #tripleS# #트리플에스# #トリプルS# #トリプルエス# #トエス#
0
11
1.3K
169
Forward to community
Spec sheets sell FLOPS. Decode spends bytes. Single-user BF16 decode can sit at roughly 1 FLOP per byte. On an H100 SXM, that puts its theoretical ceiling near 0.3% of peak math throughput.* Batching moves the point right: more tokens share each weight read. FP8 has a twist. Halving weight bytes moves the point right, but native FP8 math doubles the compute roof too. The ridge moves with you. Decode gets faster, yet the batch needed to reach the roof stays about the same. Attention limits the gains. Each user brings a KV cache. Longer contexts mean more bytes to read, so sharing weights buys less. Smaller caches and speculative decoding can help. *Teaching estimate for weight-dominated, batch-1 decode, not measured GPU utilization.
Show more
Vercel Sandbox now has memory observability. 𝚟𝚌 𝚖𝚎𝚝𝚛𝚒𝚌𝚜 𝚟𝚎𝚛𝚌𝚎𝚕.𝚜𝚊𝚗𝚍𝚋𝚘𝚡.𝚖𝚎𝚖𝚘𝚛𝚢_𝚞𝚜𝚎𝚍_𝚋𝚢𝚝𝚎𝚜 You and your agent can see P75, P95, and average usage for a single project or your entire team.
Show more
BULLISH: Solana is tripling transaction size from 1,232 to 4,096 bytes. ⚡ More complex trades 🧩 More app logic per transaction 🚀 More room to build Another step toward making $SOL more capable at scale.
Show more