惯性聚合 高效追踪和阅读你感兴趣的博客、新闻、科技资讯
阅读原文 在惯性聚合中打开

推荐订阅源

阮一峰的网络日志
阮一峰的网络日志
J
Java Code Geeks
Martin Fowler
Martin Fowler
宝玉的分享
宝玉的分享
V
Visual Studio Blog
OSCHINA 社区最新新闻
OSCHINA 社区最新新闻
M
MIT News - Artificial intelligence
U
Unit 42
博
博客园 - 三生石上(FineUI控件)
博
博客园 - 聂微东
The GitHub Blog
The GitHub Blog
I
InfoQ
WordPress大学
WordPress大学
H
Help Net Security
D
Docker
B
Blog
腾
腾讯CDC
A
About on SuperTechFans
Recent Announcements
Recent Announcements
雷峰网
雷峰网
有赞技术团队
有赞技术团队
C
Check Point Blog
Y
Y Combinator Blog
Cyber Security Advisories - MS-ISAC
Cyber Security Advisories - MS-ISAC

Hacker News - Newest: "AI"

AI can't read an investor deck AI as an attorney? Student uses ChatGPT, Gemini to sue UW over alleged racial discrimination Hacking MCP Servers in AI Systems – The Rug Pull: Tool Changes After Approval GitHub - MeepCastana/KubeezCut: Free Web based video editor Can AI judge journalism? A Thiel-backed startup says yes, even if it risks chilling whistleblowers Coming soon: 10 Things That Matter in AI Right Now DARPA built an AI to fact-check enemy weapons claims What explains heterogeneity in AI adoption? When AI Meets Muscle: Context-Aware Electrical Stimulation Promises a New Way to Guide Human Movements - Department of Computer Science AI Changed How We Build. It Did Not Change What Matters. Linux rules on using AI-generated code - Copilot is OK, but humans must take 'full responsibility for the… Meta spins up AI version of Mark Zuckerberg to engage with employees Code Mode: Let Your AI Write Programs, Not Just Call Tools | TanStack Blog GitHub - Delavalom/graft: Go framework for building AI agents. Type-safe tools, multi-provider (OpenAI, Anthropic, Gemini, Bedrock), zero vendor SDKs. India's TCS tops estimates, says new AI models did not dent services demand Gen Z's fading AI hype Strong feeling: we are in a folded AI reality GitHub - machinarii/total-recall-catalog: A reference catalog of latest knowledge retrieval, memory & RAG systems GitHub - mensfeld/code-on-incus: Give each AI agent its own isolated machine with root, Docker, and systemd. Active defense detects and stops threats automatically.. Quantization, LoRA, and the 8% Problem: Benchmarking Local LLMs for Production AI Iran war: We spoke to the man making Lego-style AI videos that experts say are powerful propaganda Powell, Bessent discussed Anthropic's Mythos AI cyber threat with major U.S. banks GitHub - immartian/bellamem: Persistent belief-graph memory for AI agents. Retrieves decisive context by importance — not recency, not RAG, not /compact. recursive-mode: The Repo-Native Operating System for AI Engineering After the attack on Sam Altman's home, will AI CEO's go on the offensive? The biggest advance in AI since the LLM Opus 4.6 vs GPT 5.4 One Prompt Unity World Generation Test “AI polls” are fake polls Client Challenge Can AI be a 'child of God'? Inside Anthropic's meeting with Christian leaders
Parametric Memory — Persistent, Verifiable Memory for AI
Parametric Memory · 2026-06-23 · via Hacker News - Newest: "AI"

Persistent, Verifiable Memory for AI

Built for developers using Claude, GPT, and any MCP-compatible agent.

Get Your InstanceSee It Live

p50 recall latency
0.045ms

Markov hit rate
64%

token savings (compact proofs)
37%

dedicated instance, starting
$5/mo

instance setup time
<60s

Capabilities

Built on verifiable mathematics.

Not approximations or fuzzy similarity search. Every claim is provable, every number is measured in production.

Merkle Proofs

Every atom is cryptographically committed.

RFC 6962-compliant SHA-256 Merkle tree. Tamper with one atom — the root hash changes. AI clients verify integrity without trusting the server.

37% token savings vs raw proofs

Markov Prediction

Your next memory is already warm.

Variable-order Markov chain predicts which atoms you'll access next. Pre-fetches context before the query arrives. Weights decay at 0.5^(days/7) — recency matters.

64% hit rate in production

Sub-ms Recall

0.045ms p50. Not a typo.

LevelDB with JumpHash sharding across 4 independent Merkle trees. Dedicated instance means zero contention. No shared cluster, no noisy neighbours.

0.045ms p50 · 1.2ms p99

MCP-Native

25+ tools. Plug in, don't integrate.

Streamable HTTP MCP transport. Works with Claude, Cowork, and any MCP-compatible client. OAuth2 + Bearer auth. The client your AI already speaks.

25 MCP tools · Streamable HTTP

Architecture

From atom to proof in four steps.

No black boxes. Every stage is deterministic, measurable, and verifiable.

  1. 01

    Atom arrives

    API call or MCP tool deposits a memory atom. Immediately hashed and committed to the Merkle tree.

  2. 02

    JumpHash routes to shard

    The atom key is deterministically routed to one of 4 LevelDB shards via JumpHash — O(1), no coordination.

  3. 03

    Markov arc fires

    The transition is recorded. The Markov predictor updates its weight table and schedules predictive prefetch.

  4. 04

    Proof returned

    On read, the atom comes with a Merkle audit path. Client verifies root hash. Tamper-evident, always.

Full architecture documentationGet Your Instance

Live Demo

Watch the memory substrate live.

Real atoms. Real Merkle rehash cascades. Real Markov arc predictions. Connect to a live MMPM instance and watch memory form in real time.

See It Live

Get started today

Dedicated substrate. Cryptographic proofs. Markov prediction.
Your own Merkle tree — not a row in someone else's database.

Get Your Instance

Starting at $5/month · 30-day money-back guarantee · Cancel anytime

Starter $5 · Solo $9 · Professional $29 · Team $79 · Enterprise Cloud $299 · Self-Hosted $499

此内容由惯性聚合(RSS阅读器)自动聚合整理，仅供阅读参考。 原文来自 — 版权归原作者所有。