speculative-execution

Spectre attack against SGX enclave

[NeurIPS'23] Speculative Decoding with Big Little Decoder

Microarchitectural exploitation and other hardware attacks.

🐉 Hail Hydra — Multi-headed speculative execution framework for Claude Code. 10 AI agents, 3x faster, ~70% cheaper. Inspired by speculative decoding.

KLEESpectre is a symbolic execution engine with speculation semantic and cache modelling

Run AI models too large for your Mac's memory — at near-full speed. Intelligent expert caching, speculative execution, and 15+ research techniques for MoE inference on Apple Silicon.

Advanced 64-bit RISC-V RVA22 soft SoC (3-way out-of-order). Scores 18 in Geekbench 5.4.1 @ 167 MHz

Variant 1 of the Spectre attack which is to bypass the bounds checks in the target process and retrieve the private data. Here in this example, I have demonstrated how to retrieve a string data using the attack.

Package speculatively provides a simple mechanism to re-execute a task in parallel only after some initial timeout has elapsed.

⚡ 100% offline, zero-cost drop-in alternative to TypeSafe Jev System One decision models in Rust (<18ms on CPU)

Synthesizes dependency DAGs across speculative agent tool calls, resolving parameter references and grouping independent invocations into parallel batch payloads | GenPark: https://www.genpark.ai | LinkedIn: https://www.llinkedin.com/company/92955550

Synthesizes dependency DAGs across speculative agent tool calls, resolving parameter references and grouping independent invocations into parallel batch payloads | GenPark: https://www.genpark.ai | LinkedIn: https://www.llinkedin.com/company/92955550

Experimental implementation of speculative model serving

High-Performance Speculative Tool-Execution Runtime for Autonomous AI Agents with Conservative Dependency & Effect Verification

open source speculation benchmarking for 2 phase protocol transactions

State-aware hedged requests for serverless GPU inference — return the first valid result and cancel the losers with an audited receipt.

Research project about Spectre attack on XiangShan open-source RISC-V processor

Read ahead of your agent. An MCP proxy that learns, from an agent's recorded tool calls, which reads come next, and serves them in the same tool result, so the agent needs fewer LLM turns.

Spectre attack variant 1 is used here to extract an image data buffer from a program. The output image data is then decoded and median filters are applied to remove the hot pixels in the output image. Here I have included files for RGB and base64 format.