2026年8月22日 技术热点总结

项目简介：Offline vanity address generator for Bitcoin and Ethereum. CPU multi-process search with interactive CLI menu. Supports Legacy, Nested SegWit, Native SegWit, Taproot, and ETH (EIP-55).

技术栈：Python

核心介绍：Offline vanity address generator for Bitcoin and Ethereum. VanityKit searches for private keys whose corresponding addresses match a user-specified pattern (prefix, suffix, or substring). All key generation and address derivation run locally. The tool makes no network requests and contains no telemetry.

关键特性：Multi-process parallel search；Live progress reporting (attempt…

项目简介：The simplest, thinnest harness that gives an LLM complete freedom to control a Mac.

技术栈：Python、accessibility、agent、automation、cdp、computer-use、macos、python

核心介绍：The simplest, thinnest harness that gives an LLM complete freedom to complete virtually any task on a Mac. The agent writes what is missing, mid-task. No framework, no recipes, no rails. One Python process connected directly to macOS, your real browser, and your files. Paste this into Codex or Claude Code:

项目数据：⭐ 685 Stars，🍴 45 Forks

项目简介：A 3D fruit fly living on your macOS desktop, driven by a live spiking simulation of the real FlyWire connectome

技术栈：Swift

核心介绍：DesktopFly 🪰 A 3D fruit fly that lives on your macOS desktop — driven by a live spiking simulation of the real FlyWire connectome. It walks across your windows, grooms, sleeps, and decides to flee your cursor with the same neurons a real fly uses. The fly’s brain window: 23,210 real neuron soma positions from FlyWire v783, with live spikes flashing at real neuron locations. The two glowing yellow markers are the Giant Fibers — the escape command…

项目简介：An inference chip design that runs a real LLM (Qwen2.5-0.5B) on FPGA — one transformer decoder layer in RTL, every silicon value bit-exact against a golden model. 0.56 tok/s measured, a 140× climb, full evidence trail.

技术栈：Python

核心介绍：alt=”APEX — tinyNPU: the attention engine that remembers”> decoder layer in real RTL — attention, KV-cache compression, softmax, RMSNorm, RoPE, SwiGLU, residual — every block bit-exact against an executable golden model, running real Qwen models through the verified pipeline, and brought up on real FPGA hardware. Pre-silicon, and every numbe…

技术栈：TypeScript

核心介绍：Deterministic geometric blobatars from any string. No dependencies, ~4.4 KB bun add blobatar # npm / pnpm / yarn all work too

项目数据：⭐ 655 Stars，🍴 39 Forks

项目简介：SVG recreation of the x.ai bot avatar. One shape morphing through 14 states, measured off the reference video frame by frame.

技术栈：TypeScript、animation、avatar、morphing、svg、svg-animation、typescript、vue

核心介绍：An SVG recreation of the x.ai bot avatar: one filled black shape that morphs between 14 states, two white shapes for the eyes that morph independently, on a plain background. No animation library. Then open http://localhost:5190. pnpm test # vitest pnpm build # vue-tsc –noEmit && vite build Vue 3, Vite, TypeScript, Tailwind 4. No ESLint and no Prettier: v…

项目简介：Offline utilities for Coldcard hardware wallet users: PSBT inspection, BIP39/dice entropy, Seed XOR split/combine, BBQr encode/decode, output descriptors, and firmware verification guidance. Companion to official Coldcard firmware. Not affiliated with Coinkite.

技术栈：Python、airgap、airgap-devkit、airgap-download、airgap-setup、airgap-tutorial、bitcoin、bitcoin-wallet、coldcard

核心介绍：Offline utilities for Coldcard hardware wallet users. Companion to the official Coldcard firmware and ckcc-protocol. All core tools run locally. Optional silent stats ping (one POST, errors ignored) doe…

Large language model agents can adapt to complex tasks by constructing workflows at inference time, but procedures discovered in one episode are usually discarded after execution. Existing skill libraries provide reusable executable routines, but are typically assembled offline and do not grow from the agent's own workflows. We introduce FlowEvo, a training-free framework in which workflows and skills co-evolve at inference time. FlowEvo compiles successful workflows into callable skills, stores them in a persistent bank, and uses retrieved skills either through direct execution or as conte…

We introduce TinyCast, an attention-free zero-shot forecaster that emits a predictive distribution from 146,505 parameters, on the premise that at this size the periodic structure of a context is worth computing rather than learning. A zero-parameter spectral detector supplies the dominant periods, the context is folded on their phase, and a dilated convolutional encoder and a block-autoregressive quantile decoder model the rest. It is smaller than every zero-shot entry on the GIFT-Eval board whose parameter count can be established. On probabilistic accuracy it defines the size-accuracy fr…

Should you replace your text-embedding pipeline with a large language model? We answer this with a controlled, cost-aware comparison of ten LLMs across six families and 26 embedding models (118M to 14B parameters) on 37 tasks spanning classification, semantic textual similarity (STS), clustering, pair classification, and retrieval. In aggregate the two paradigms are effectively tied: the best LLM (Gemini 3.1 Pro, 77.6) and the best embedding model (77.2) differ by 0.4 points. Their strengths differ by task: LLMs lead on reasoning-heavy retrieval, embedding models lead on classification, and…

Long-horizon robot manipulation requires a robot to both execute individual skills reliably and sequence them coherently over extended tasks. Most hierarchical vision-language-action (VLA) models make each such decision with a single forward pass, leaving no mechanism to allocate additional computation to difficult or consequential choices. We introduce τ_0-VLA, a hierarchical robot foundation model that formulates high-level subtask generation as a compute-scalable inference problem through world-model-guided test-time computation. At each inference step, the high-level policy uses executi…

LLM coding agents issue Bash commands through interfaces that may serialize, wrap, and reparse model output. Matched execution scores alone cannot distinguish command-generation errors from failures introduced after generation. QuoteBench measures this boundary with exact final-state validation on 56 one-shot tasks from 14 incident-derived families, crossing the generation contract with the execution transport around one deliberately unescaped added parser. Escaping at the interpolation point reproduces each replayed reply's raw-path outcome, so any recovery under a disclosed boundary must …

Modern LLM agents are often improved by modifying prompts, tools, or workflows manually, while the executable scaffold surrounding the model—the harness—is typically treated as a fixed artifact after deployment. This work studies an alternative where the harness is task-specific and continuously evolvable: each task family maintains its own harness, which is hot-swapped across iterations through a fixed task-injection seam and rewritten using environment feedback. We introduce Hierarchical Self-Improvement (HSI), a framework in which a single frozen LLM M operates across three hierarchi…