标签：#ens

Introduction Modern AI systems rarely rely on a single model anymore. A fraud detection pipeline might combine specialists for: Transaction analysis Identity verification Device fingerprinting Network analysis Similarly, RAG pipelines, LangGraph workflows, and other multi-agent systems often have several AI agents collaborating before producing a final decision. As these systems become more complex, one question becomes surprisingly difficult to answer: Which agent actually influenced the final decision? Running four or five agents doesn't necessarily mean all of them contributed. Sometimes a single specialist completely determines the outcome while the rest simply add latency and compute cost. Most multi-agent frameworks make it easy to build agent workflows—but they don't tell you which agents actually mattered . That question led me to build agent-ablation , a lightweight TypeScript library for performing leave-one-out ablation testing on multi-agent decision systems. Why I built this While experimenting with multi-agent systems, I kept asking myself questions like: Which specialist actually changed the final verdict? Which agents consistently influence decisions? Are some agents effectively redundant? Am I paying for LLM calls that never affect the outcome? Answering those questions usually meant manually removing agents, rerunning experiments, and comparing outputs. That quickly became tedious. I wanted a simple utility that could automate this experiment. Instead of guessing which agents mattered, I wanted to measure their influence. That's why I built agent-ablation . The Idea The core algorithm is intentionally simple. Given a set of agent findings and a deterministic decision function: Compute the baseline decision. Remove one agent's finding. Recompute the decision. Compare the new verdict with the baseline. Repeat for every agent. If removing an agent changes the verdict, that agent is load-bearing . Otherwise, it wasn't necessary for producing that parti

Agents can already talk. MCP and A2A exist. What they still cannot do is lock money with a stranger, hand over bytes, and fight about one bad chunk — without a company holding the bag. That gap is what ArthNeura is for. Two repos on purpose arthneura-core is a Substrate solo-chain. pallet-agent-registry — ML-DSA-65 DID, deposit, reputation pallet-vector-db — Merkle commitment, dispute bound to one chunk index pallet-escrow — lock / release / refund Pallets do not import each other. The runtime wires traits. arthneura-market is only discovery. Listings, signed offers, delivery URLs. No keys. No funds. No verdict. The board names the next chain call and does not submit it. Status Pre-testnet. v0.1. Local --dev node. Not a public network. Not a token post. https://github.com/arthneura/arthneura-core https://github.com/arthneura/arthneura-market https://github.com/arthneura

“Store the data” is only the beginning of the problem. The difficult questions usually come afterward: What structure does this data actually have? Which fields are missing or inconsistent? Which values are invalid? Which changes are safe to apply automatically? What exactly changed after a repair? Can the system prove that its storage and indexes are still consistent? I built Atlas to answer those questions inside the database engine itself. Atlas is a zero-dependency embedded database for semi-structured data. It stores records, builds a full-text search index, infers schema, analyzes data quality, proposes safe repairs, preserves uncertain records, and records an audit trail of applied changes. It does not use SQLite or SQL. It is not intended to replace SQLite for relational workloads. Instead, Atlas focuses on a gap that is usually handled by external scripts and tools: Data inspection, diagnosis, and safe repair as first-class database capabilities. That is the problem Atlas was built to solve. Why data quality belongs inside the database engine Most databases are very good at storing and retrieving data. That is necessary, but real-world data work rarely stops there.** Operational records, imported JSON, CSV files, event payloads, and semi-structured documents often arrive with problems: { "id" : "T-1" , "title" : " Connection timeout " , "priority" : "HIGH" } { "id" : "T-1" , "title" : "connection timeout" , "priority" : "high" } { "id" : "T-2" , "title" : "Unicode café search" , "priority" : null } These records contain several potential issues: Duplicate logical identifiers Leading or trailing whitespace Inconsistent capitalization Null-like values Missing fields Mixed data types Malformed email addresses Different date formats Inconsistent structures across records A storage engine can preserve these values perfectly while still leaving the data difficult to understand and use. The usual response is to add external tools: A schema profiler A data-quality

Somewhere in the DataLens build, my teammate and I hit the wall every "zero-dependency" project eventually hits: the anomaly detector needed a neural net, and the rulebook said no third-party packages. No NumPy. No pandas. No scikit-learn. Just Python 3.14's standard library. Our first reaction was denial. You cannot build an ANN without a matrix library — everyone knows that. numpy.dot() is basically load-bearing infrastructure for machine learning in Python. We spent an embarrassing amount of time trying to convince ourselves some obscure math submodule secretly did vectorized linear algebra. It doesn't. There is no shortcut. If you want matrix multiplication in pure stdlib Python, you write nested for loops and you like it. What we normally would have installed In any other project, this is a two-second decision: pip install numpy , import it, move on with your life. Matrix ops, broadcasting, vectorized activation functions — all free. Neither of us had ever really had to think about how A @ B works under the hood, because neither of us had ever had to write it ourselves. What it actually took to replace it An autoencoder needs: matrix multiplication, transpose, element-wise activation functions (sigmoid, ReLU), and gradient computation for backprop. Without NumPy, every one of those is a hand-rolled function operating on nested Python lists. Matrix multiply becomes three nested loops instead of one line. A forward pass that would be a single .dot() call turns into a small file of helper functions: matmul() , transpose() , add_bias() , sigmoid() , sigmoid_derivative() . We split it — one of us built the forward pass and activation functions, the other took backprop and the training loop — and then spent a good while debugging the seam where the two met. The genuinely hard part wasn't the math — it was performance. Pure Python loops over lists of lists are slow, and profiling a dataset with a few thousand rows through even a small autoencoder made that obvious fas

Longshot is a Firefox screenshot extension I wrote to replace FireShot: full page, visible area, drag region and element capture, an editor with eleven annotation tools, export to PNG, JPEG, WebP and PDF, and local OCR that produces a searchable text layer. It has no runtime dependencies. The code is not public, so this is a description rather than an invitation to read it. At one point it had 130 passing Node assertions across six suites, zero failing. Printing could not open a dialog at all. Not "printed the wrong thing". The print command hung indefinitely and no dialog ever appeared. The suites did not go amber, or flake, or report a warning. They reported 130 passed, 0 failed, which is what they had reported the day before and what they would have gone on reporting. Why nothing caught it printCanvas encoded each slice of the image to a blob URL and awaited img.decode() . That call does not resolve for an image inside a display:none subtree, and the print stylesheet creates exactly such a subtree by design, since the container has to be hidden on screen. So the await never returned, and the dialog never opened. Every part of that failure is a meeting point between my code and the browser: the decode promise's behaviour, the stylesheet's effect on the subtree, and the ordering between them. None of it is reachable by a function you can call from Node. The six suites test band arithmetic, canvas dimension limits, filename sanitising, the background module graph under stubbed extension APIs, PDF structure and scan geometry. All of that is worth testing and none of it goes near a real DOM. The second bug in the same batch has the same shape one level in. Choosing PDF broke "Open in editor", because deliver() handed the editor the PDF blob and createImageBitmap cannot decode one. That is not a browser boundary; it is one internal stage handing another something it cannot accept. Both stages were tested; the seam between them was not. That is the pattern worth naming.

💻 One thing I dislike about coding-agent setups is how quickly they become part of one specific machine. Provider config goes in one place, session state somewhere else, local models live in another directory, and suddenly moving to a second machine means rebuilding the environment. OpenClaude-Portable takes a much cleaner approach. It packages the coding agent, runtime and persistent data into a self-contained folder. It supports cloud and local models in the same setup The project currently supports 9 provider options: Anthropic Claude OpenAI Google Gemini DeepSeek OpenRouter NVIDIA NIM Ollama LM Studio custom OpenAI-compatible APIs I like this because the portable part is not tied to one model vendor. I can use a cloud model when I want the strongest hosted option, then switch to Ollama or LM Studio when I want a local workflow. The important caveat is simple: cloud providers still need internet. Ollama can run offline after the initial setup. The "zero footprint" idea is more useful than it sounds The project redirects its persistent data into a local data folder. That includes provider settings, API keys, logs, session history, agent memory and local Ollama files. According to the repository, it does not write configuration into the host system. For me, this is the real feature. I do not care that the agent happens to be on a USB drive. I care that I can move the folder and keep my environment with it. 💾 There are two very different ways to run the agent The launcher offers a normal mode that asks before file writes or shell commands. There is also an optional Limitless mode that can run without approval prompts. I like that these are explicit choices rather than one hidden permission switch. For normal development I would keep approval mode on. For a disposable test project or a controlled autonomous task, the second mode could be useful. Sessions can survive the move Another practical detail is session resume. The project stores session history inside the por

Faker Doesn't Know Your Entities Are Related, So I Built Something That Does You've added a second entity to the schema, wired up a @ManyToOne , and gone back to your seed script to generate fifty more rows. Ninety seconds later, the app refuses to start: unique constraint violation, somewhere inside a loop you wrote three weeks ago at 11pm. You fix it. You restart. A different field breaks a different constraint. This is the exact moment every Spring Boot developer eventually meets the real limit of tools like Faker. They're brilliant at generating a name, an email, an address. They have no idea the Payment sitting in front of them needs a Counterparty to already exist. So you do what everyone does: hand-write the wiring. Create parents first. Hold onto their generated IDs. Wire them into children. Hope you didn't just violate a @NotNull somewhere in the process. It works, for a while. Then the schema changes, and the script quietly stops matching reality until the next 3am debugging session finds out the hard way. I hit this enough times that I stopped patching the script and looked at the actual problem: the information needed to seed this correctly already exists. It's sitting right there in the entity, in the annotations you already wrote. @ManyToOne , @NotNull , @Column(unique = true) , JPA already knows the shape of your data. Nothing should need to be told that twice. That became SynthForge . The core idea Instead of writing a script that generates data, you annotate the entity: @Entity @Seed ( count = 50 ) public class Counterparty { /* fields only */ } @Entity @Seed ( count = 200 ) public class Payment { @ManyToOne ( optional = false ) private Counterparty counterparty ; } Start the app in a dev profile. Both tables populate, correctly ordered, on every restart. No seed method. No calling code, anywhere. The entity is the seed script. What's actually happening underneath Entity scanning. SynthForge reads JPA-managed attributes through the jakarta.persisten

An approval rule becomes useful when you can test what happens on both sides of it: the forbidden action is refused, and the permitted decision leaves evidence. A happy-path demo alone cannot show that distinction. Here is a runnable example using Accordo, the open-source framework coding agents use to build custom CRMs. A synthetic customer wants 30 seats of an Enterprise Plan and requests 25% off. The existing policy permits automatic approval through 10%; above that, through 50%, it requires a user decision. Run it locally You need Git, Node.js 22.16 or newer, npm, and internet access for cloning and dependency installation. Start in an empty working directory: git clone https://github.com/khaoss85/agent-crm.git framework-source cd framework-source git checkout 3b5b5f0c4c3e582e48d54501136024b064756daa node --no-warnings examples/recipes/quote-approval/run.mjs ../my-quote-crm The pinned recipe source creates a project, installs its dependencies and composes the existing commercial package. It then starts a temporary server on localhost and drives the public SDK through HTTP. The catalog is a fixture; the business journey does not call an external provider. It uses source from the checkout, independently of the npm scaffolder release. Check the refusal, then the decision The script contains assertions for each transition: Server pricing produces EUR 3,750 once and EUR 2,400 per month after discount. These are synthetic quote amounts, kept in separate periods. Submission under policy version 1 freezes a commercial snapshot and enters pending_approval . An approval request from the simulated agent receives HTTP 403 with HUMAN_APPROVAL_REQUIRED . The quote and approval remain pending, and no business audit entry is added. A simulated user approves. The quote becomes approved , with one user decision audit and a completed trace. The submitted snapshot remains unchanged. There is one quote version and one approval record. The refusal also has a failed trace. That is a u

The rise of AI has brought an avalanche of new terms and slang. Here is a glossary with definitions of some of the most important words and phrases you might encounter.

I build text-to-SQL agents on Oracle and Postgres for a living. Every one of them had the same bug, and it wasn’t in my code. It was in the order of operations. The bug The schema goes into the prompt before the query runs. Row-level security runs when the query runs. So the model sees a table the user can’t read, writes perfectly valid SQL against it, the database returns zero rows, and the agent says “no records found”. A wrong answer, delivered with confidence. Vanna (23k stars, archived March 2026) applied identity exactly there: at execution, after the model had seen everything. The fix Apply identity at selection. Decide which tables the model is shown, per caller, before any SQL exists. A restricted table isn’t ranked low — it’s absent. from schemagate import Catalog, Principal cat = Catalog().bootstrap("postgresql://localhost/app") cat.restrict("hr_compensation", roles=["payroll"]) analyst = Principal("okta:jdoe", roles={"analyst"}) cat.select("salary by employee", principal=analyst).table_names # no hr_compensation pip install schemagate — one dependency, no API key, any SQLAlchemy database. The side effect that pays for it You’re now sending ~6 tables instead of the schema dump. Measured on the test schemas: 65–79% fewer prompt tokens on small ones, 97% on a 260-object one (16,095 → 444 per question). The selector never calls a model — BM25 plus a hashed embedder, offline, milliseconds. What broke while building it Six invented schemas found ten bugs before release. My favourite: a three-column orders_bkp outranked the real orders table, because short documents win cosine similarity. Backup and staging copies now rank below the object they shadow. The full list is in TESTING.md. Where it plugs in MCP server for Claude Desktop and Cursor, a LangChain retriever, a native Oracle 23ai VECTOR store, and a browser demo that needs no install: https://ashishsinha1602.github.io/schemagate/ Repo: https://github.com/ashishsinha1602/schemagate — tell me where it break

Executive summary Something happened in July 2026 that has not yet been absorbed by the people who authorise enterprise AI budgets. Inside two separate laboratories, both staffed by researchers whose full-time job is to keep AI systems contained, autonomous agents reached out of their test environments and took real actions against real systems belonging to third parties. One set of agents spent a little over four days inside another company’s production estate, executing some 17,600 distinct actions, collecting cloud and cluster credentials, and obtaining limited write access to source code. Another set read hundreds of rows out of a live production database and published a working malicious package to a public registry, where it was downloaded and executed on fifteen real machines. Neither event was a jailbreak in the cinematic sense. There was no clever exploit of a hardened perimeter. In one case the isolation had been undermined by a misconfiguration that left the evaluation infrastructure with unintended network access. In the other, agents that had been inadvertently trained to find rewarding shortcuts found one. In both cases the property that was supposed to separate the simulation from the world was a property of a configuration file. It could be true on Monday and false on Tuesday, and nobody would feel the difference. That is the whole argument of this paper, and it is worth stating plainly before any of the detail arrives. The organisations that lost control of their agents were not careless. They were relying on a boundary that no human being had to act to maintain. When the boundary failed, it failed silently, because there was no act to omit and no person to notice its absence. An air gap is a claim about topology. It is asserted once and inherited forever. Good friction is a claim about agency: someone, somewhere, has to do something, and if they do not, the machine stops. Enterprises are about to run this experiment at industrial scale. Deloitte’s

GitHub热门项目 | The open source AI research agent. | Stars: 8,904 | 262 stars this week | 语言: TypeScript

GitHub热门项目 | The free and privacy-friendly screen recorder with no limits 🎥 | Stars: 18,664 | 86 stars this week | 语言: JavaScript

GitHub热门项目 | Node.js & React bindings for Zed’s GPUI. Build memory efficient native apps with React and no Electron | Stars: 1,631 | 76 stars today | 语言: Rust

GitHub热门项目 | A free + OSS logo generator powered by Flux on Together AI | Stars: 8,593 | 111 stars today | 语言: TypeScript

GitHub热门项目 | Standalone iOS app to spoof GPS location without jailbreak. Includes Shadowrocket/Surge/Loon/QX/Stash module. | Stars: 3,856 | 20 stars today | 语言: JavaScript

GitHub热门项目 | Claude-first paid-media operations skill for Claude Code across 12 ad platforms (Google, Meta, YouTube, LinkedIn, TikTok, Microsoft, Apple, Amazon, Reddit, Pinterest, Snapchat, X): source-grounded audits, deterministic scoring, versioned JSON reports, and capability-gated account changes. | Stars: 8,948 | 94 stars today | 语言: Python

GitHub热门项目 | Production-grade MCP server giving Claude 27 security intelligence tools across 21 APIs — CVE lookup, EPSS scoring, CISA KEV, MITRE ATT&CK, Shodan, VirusTotal, and more. | Stars: 1,458 | 46 stars today | 语言: Python

GitHub热门项目 | 本项目采用 CC BY-NC-SA 协议，禁止任何商业化行为，任何衍生项目必须保留本项目地址并以相同协议开源 | Stars: 9,585 | 171 stars today | 语言: TypeScript

Originally published on lockmargin.com Last weekend I asked myself an uncomfortable question: could someone post my app and a valid license key on a torrent, and anyone use it for free? The answer is yes. That question sat with me all weekend. I couldn't build anything else until I mapped it, so this is the map. Context: I'm building LockMargin, a local-first invoicing app for freelancers. $49 once. No account. No telemetry. No phone-home of any kind. The license is a signed token the app verifies locally. The threat model First, I narrowed the problem. I'm not trying to stop professional cracking crews or commercial resellers. I'm trying to survive the casual kind: a key shared in a chat, an installer plus key on a torrent, "send me your key" between friends. Then I drew a hard line. If a mechanism breaks on a new laptop, a wiped disk, a VM - it's out. If it needs a server - out. If it needs an account - out. The app has to keep working even if my company disappears. Every option on my list had to pass that test. The honest ceiling Here is the fact most DRM marketing hides: if an application can locally answer "this token is valid," then anyone holding the same token gets the same answer. Copying the authorization is always possible when verification is local. That is not an implementation flaw. It is the architecture. Two different things get conflated here. Authenticity: did I publish this binary? Code signing answers that. Authorization: does this person own a license? A signed local token answers that. Neither prevents a buyer from handing a valid token to somebody else. And two different threats. Sharing is when a buyer gives their key to another person - the contractual and social layers help with that. Cloning is when the installer and a valid key travel together. That is where the hard limit appears: if a real installer plus a real key spreads, an offline application cannot tell the new user from the owner. I say that plainly, because pretending otherwise i