标签：#t

Optimizing RAG at Scale: Chunking, Retrieval, and the Bayesian Search That Cut Latency 40% How we moved from "semantic search + hope" to a measured, tunable retrieval pipeline with 95% recall@10 The RAG Reality Check Everyone ships RAG the same way: chunk by 512 tokens, embed with text-embedding-3-small , top-k=5, stuff into context. It works for demos. Then you hit production: Legal contracts: 512 tokens splits clauses mid-sentence API docs: 1000-token chunks drown signal in noise Customer tickets: Conversational context needs overlap, not fixed windows Latency: 500ms embedding + 200ms vector search + 300ms LLM = 1s+ per query We rebuilt our retrieval layer from first principles. Here's what actually moves metrics. Chunking: One Size Fits None # rag/chunking.py from abc import ABC , abstractmethod from dataclasses import dataclass @dataclass class Chunk : text : str metadata : dict token_count : int chunk_id : str class ChunkingStrategy ( ABC ): @abstractmethod def chunk ( self , document : str , metadata : dict ) -> list [ Chunk ]: ... class FixedTokenChunker ( ChunkingStrategy ): """ Baseline. Good for homogeneous content. """ def __init__ ( self , chunk_size = 512 , overlap = 50 ): self . chunk_size = chunk_size self . overlap = overlap class RecursiveChunker ( ChunkingStrategy ): """ Respects structure: markdown headers, code blocks, paragraphs. """ def __init__ ( self , separators = [ " \n ## " , " \n ### " , " \n\n " , " \n " , " " ], chunk_size = 512 ): self . separators = separators self . chunk_size = chunk_size class SemanticChunker ( ChunkingStrategy ): """ Uses embedding similarity to find natural boundaries. """ def __init__ ( self , model = " text-embedding-3-small " , threshold = 0.7 ): self . model = model self . threshold = threshold class AgenticChunker ( ChunkingStrategy ): """ LLM decides boundaries. Expensive but highest quality for complex docs. """ def __init__ ( self , model = " gpt-4o-mini " ): self . model = model Our production config by

Building Production-Grade LLM Evaluation Pipelines: From Vibes to Metrics How we replaced "looks good to me" with automated evaluation catching 92% of hallucinations before deployment The Problem: Why "Vibe Checks" Fail in Production Three months ago, our team shipped a RAG-based customer support assistant. It worked great in testing — we'd ask it questions, read the answers, and say "yeah, that looks right." Then it hit production. A customer asked about their billing cycle. The assistant confidently cited a policy that didn't exist. Another asked about API rate limits and got numbers from a competitor's documentation. By the time we caught it, 500+ users had seen hallucinated responses. The post-mortem was brutal: we had zero automated evaluation . Our test process was literally "ask 5 questions, read answers, thumbs up." What Production Evaluation Actually Needs Academic benchmarks (MMLU, HellaSwag) don't tell you if your system works for your use case. Production evaluation needs: Domain-specific judges — Your criteria, not generic "helpfulness" Speed — Evaluation must run in CI/CD, not overnight Regression detection — Know immediately when a prompt change breaks things CI/CD integration — Block merges that degrade quality Golden dataset management — Versioned, stratified, growing test cases Architecture: The Evaluation Pipeline ┌─────────────┐ ┌──────────────┐ ┌────────────────────┐ ┌──────────────┐ │ Test Cases │────▶│ LLM Under │────▶│ Judge Ensemble │────▶│ Metrics & │ │ (Golden Set)│ │ Test │ │ - Faithfulness │ │ Regression │ └─────────────┘ └──────────────┘ │ - Instruction F. │ │ Detection │ │ - JSON Schema │ └──────┬───────┘ │ - Custom LLM │ ▼ └────────────────────┘ ┌──────────────┐ │ Dashboard/ │ │ PR Comments │ └──────────────┘ Core Abstractions # eval/base.py @dataclass ( frozen = True ) class TestCase : id : str input : dict [ str , Any ] expected : dict [ str , Any ] | None = None tags : list [ str ] = field ( default_factory = list ) # ["edge-case",

The front-running tax that bled crypto for a decade needs only observable intent and a party that controls order. Agent marketplaces are rebuilding both. In September 2020, a security researcher who goes by samczsun found about $12 million of someone else's cryptocurrency sitting in a vulnerable contract, exposed, and realized he had a few minutes to rescue it before someone less friendly noticed. He wrote the rescue transaction. Then he stopped, because he understood the problem with sending it. The moment his transaction hit Ethereum's public waiting area, the mempool, every bot watching that space would see a profitable move spelled out in plain code, copy it, pay a higher fee to jump ahead of him, and take the $12 million themselves. His rescue would become their heist, and he would have personally handed them the map. He wrote about this later in an essay called "Escaping the Dark Forest," borrowing a metaphor from Dan Robinson and Georgios Konstantopoulos at Paradigm, who had borrowed it from Liu Cixin's science fiction: an environment where any signal of your presence gets you killed, so the only survivors are the ones who stay silent and shoot first. The mempool is a dark forest. Broadcasting a valuable intention into it is detection, and detection is death. Samczsun survived only by refusing to play the open game. He submitted his rescue privately, straight to a miner, bypassing the public mempool entirely, so the predators never saw it coming. That story is usually told as a piece of crypto lore. I want to tell it as something else, because the thing that killed transactions in the dark forest was never really about blockchains. It was about a shape, and that shape is quietly being rebuilt inside the AI agent marketplaces that a lot of people are racing to launch right now. When it finishes, the same predators will be back, and this time the prey will be your agents. The three conditions, and why blockchain was just the extreme case The phenomenon samczsun

Alphabet, Google's parent company, is reportedly working on a new chip designed to make its Gemini models run much more efficiently.

Publicado originalmente en Justicia Liberal . El hecho y su lectura económica La reconfiguración de lealtades internas en el bloque peronista del Congreso que reporta LA17 podría leerse como un episodio más del folletín peronista: disputas de conducción, alineamientos con gobernadores, señales cruzadas hacia 2027. Pero desde una perspectiva económica aplicada, el fenómeno tiene consecuencias concretas sobre variables que le importan a cualquier empresa, ahorrista o asalariado argentino. La inestabilidad legislativa no es un dato político neutro. Es un factor de riesgo que los mercados descuentan en tiempo real. Incertidumbre legislativa y su precio en variables macro Cuando un bloque opositor mayoritario se fragmenta, el resultado inmediato no es la debilidad del peronismo: es la imprevisibilidad del Congreso. Y la imprevisibilidad tiene precio. El riesgo país argentino, que según datos del BCRA y operadores de mercado secundario rondó los 600-700 puntos básicos durante buena parte de 2025, es en parte una prima por incertidumbre institucional. Cada vez que el Poder Legislativo se convierte en un tablero de negociaciones opacas —donde un artículo fiscal puede ser bloqueado, modificado o aprobado según quién necesite qué favor de quién— la tasa de descuento que aplican los inversores sobre activos argentinos sube. Eso se traduce en mayor costo de financiamiento para el Tesoro y, por efecto derrame, para el crédito privado. El mecanismo es simple: si no sabés qué va a salir del Congreso la semana próxima, no invertís a largo plazo. Y si no invertís, no generás empleo formal ni capacidad productiva. El déficit como rehén de la interna El equilibrio fiscal que el gobierno de Javier Milei logró sostener durante 2024 —el primer superávit financiero en más de una década, según datos del Ministerio de Economía— depende en parte de que el Congreso no apruebe gastos que el Ejecutivo no puede financiar sin emisión. Ahí es donde la fragmentación peronista se vuelve peligrosa de

What Happened On July 20, 2026, cybersecurity firm Group-IB disclosed a new espionage implant dubbed HollowGraph that hijacks compromised Microsoft 365 mailboxes to run a command-and-control (C2) channel hidden inside calendar events. The malware attaches encrypted files to calendar entries dated May 13, 2050 — far enough in the future that a mailbox owner would never scroll to them — and retrieves operator instructions from the same dead drop. All traffic moves through the Microsoft Graph API, making the activity indistinguishable from legitimate M365 usage. At least 12 systems have been infected, with three actively communicating with the threat actor between June 3 and July 9, 2026. The indicators point to a targeted espionage campaign focused on Israeli organizations . Technical Analysis HollowGraph is a lightweight .NET DLL that supports only two commands: GET and SEND . To receive tasking, it queries the compromised mailbox's calendar for an event titled in the format "Event ID: <7-char-taskID>", downloads the attached file, and decrypts it using RSA and AES-256-GCM. To exfiltrate data, the implant creates a new calendar entry titled "Boss{..}ID{..}" and uploads stolen files encrypted with the attacker's public RSA key. The Group-IB research team described the mailbox calendar as a "covert dead-drop," with HollowGraph retrieving commands from events scheduled within a fixed one-hour window between 22:00 and 23:00 UTC on the far-future date. The hybrid encryption scheme uses separate RSA key pairs for inbound and outbound channels, keeping them cryptographically isolated. A second, unencrypted channel runs over DNS tunneling . HollowGraph refreshes its Microsoft Entra ID (Azure AD) credentials by querying IPv6 AAAA records from the attacker-controlled domain cloudlanecdn[.]com . Each returned IPv6 address yields 14 usable payload bytes, which the malware assembles and decodes as UTF-8 text to update its logAzure.txt configuration file — a file masquerading as a

Judge grants restraining order, saying merger "likely to violate antitrust laws."

Directly added to your Samsung Wallet account, it’s yet another cash-back credit card, this time tailored for Samsung stans.

Index funds are touted as one of the safest ways to invest. Rather than picking and choosing individual stocks, index funds let you bet on the market as a whole. So what happens when a company like SpaceX - a giant gamble, and, in my opinion, terribly overpriced - is fast-tracked into the Nasdaq-100? Does […]

The agency stressed that the July 17 voluntary recall from Taylor Farms still stands. The supplier has been linked to the ongoing cyclospora outbreak.

The aptly-named “2K Megahits 2026 Bundle” from Humble includes 15 Steam games for $15. It’s a smattering of the publisher 2K’s biggest games of the 2010s, including BioShock Infinite, Risk of Rain 2, Borderlands 3, and XCOM: Enemy Unknown. The bundle will be available until August 7th, though the bundle page notes that keys for […]

You probably know the signature PlayStation boot sound. Did you know that it's technically a multi-part chime? There's the synthy section where "Sony Computer Entertainment" shows onscreen with a white background. You only get to the next screen with the echo-y chimes and the color-filled PlayStation logo if your console recognizes the disc. How badly […]

The latest estimates reveal that the majority of greenhouse gas emissions from large-scale events such as the FIFA World Cup stem from spectators traveling to host cities.

AT&T suffers setback but will keep asking court and FCC to preempt state rules.

The end of the Ford F-150 Lightning was also the start of a new era for the automaker. After failing to capture the market with the electric version of its popular F-Series truck, Ford is going back to the drawing board with its Universal Electric Vehicle Platform. The first offering will be a $30,000 electric […]

The one-year-old startup aims to reinvent financial architecture for autonomous AI transactions.

A recent study suggests that not getting enough sleep increases the risk of weight gain, and thus of conditions like heart disease and type 2 diabetes.

Netflix has teamed up with a company called SignUp Media to offer ASL interpretations on a bunch of kids shows and movies.

Every AI writing tool does the same thing: you type a prompt, it spits out text, and you copy-paste it into Word. Then you spend the next 30 minutes manually applying headings, fixing list styles, rebuilding tables, and cleaning up formatting that broke on paste. Co-pilot lives inside Word, sure, but it's a writing assistant, not a formatting one. It drafts text, rewrites paragraphs. What it doesn't do is look at your messy 40-page report and restructure it with your actual Word styles, proper heading hierarchy, and correctly formatted tables. That's the gap we set out to close with Stylifyword. What Stylifyword Actually Does: Stylifyword is a Microsoft Word add-in paired with a companion desktop app. You give it a document, a messy draft, a copy pasted ChatGPT output, a stitched together report from five different authors and voila, it writes & edits your text when you ask it to (drafting, rewriting, summarizing the usual AI stuff). Also formats the entire document with your headings, lists, tables, and all. Outputs every change as a tracked redline in Word's Review tab. Nothing touches your document until you accept it. That third point is the key differentiator. It's non-destructive. The AI proposes, you dispose. Why Local-AI support Matters: Here's the thing that got me building this: most professionals who need AI the most can't use cloud AI tools. Corporate attorneys can't paste M&A deal terms into ChatGPT. Healthcare teams can't upload patient data to a cloud endpoint. Defense contractors, compliance officers, financial analysts, they're all locked out of the AI productivity wave because of legitimate data sovereignty concerns. Stylifyword's default mode runs entirely on your machine: 100% on-device inference via a companion desktop app. No Ollama, no command line, no model downloads to configure. Works offline, airplane mode, air-gapped networks, whatever. Zero data leaves your device, ever. Install the app, open Word, and it works. Three Ways to Run It: Not

This week's Java roundup for July 13th, 2026, features news highlighting: a reintroduction of Value Objects (Preview); the GA release of WildFly 41; the July 2026 edition of Open Liberty 26.0.0.7; point releases of TornadoVM, Apache TomEE, Java Operator SDK and LangChain4j; a maintenance release of Micronaut; a new extension, Quarkus Shim; and a new Oracle AI Agent Studio for Fusion Applications. By Michael Redlich