标签：#ai

I'm building LOOM — a small open-source language that is a machine-checked trust layer for AI-written code. I don't write it by hand anymore: an organism I built grows it, day and night, on my own machine. This is Day 6, and the whole day went to one thing — WebAssembly . Why this was a real test LOOM already runs three ways: an interpreter, and backends that compile checked code to Python and JavaScript. The thesis is "trust survives translation" — effects and provenance, proven once, hold the same on every target. WebAssembly is the strongest test of that: a low-level stack machine with linear memory, nothing like Python or JS. And there was a constraint. This machine's clang has no wasm target, and I install nothing paid or heavy. So I don't compile to wasm through a toolchain — I emit the wasm bytes myself (LEB128, the type / function / memory / global / export / code sections, the i32 stack machine) and run them through node's built-in WebAssembly . Zero dependencies. From fib to a value runtime, in a day Every step was prototyped and proven (wasm output == interpreter output) before it touched the kernel: The integer core — arithmetic, comparison, if , first-order calls and recursion. fib(10) becomes 61 bytes of real WebAssembly and returns 55, identically on the interpreter, Python, Node and wasm. A value runtime — let and integer lists in a real linear-memory heap (a bump pointer + a $cons cell allocator; head / tail are i32.load , empty is i32.eqz ). A list sums and folds by recursion, inside wasm. Sum types — (variant Tag e) becomes a tagged cell [tag-id | payload] ; match loads the tag, compares, binds the payload, branches. You can watch it: the live playground has a Compile → WAT button and WASM · fib / list-sum / match examples. Type a program, see it become real assembly, in your browser. Honest scope: ints, let , integer lists and sum types compile to wasm today. Records, closures and effects are the next frontiers (closures are the hard one — a func

TMX: The open standard AI agent memory has been waiting for The problem no one talks about: your agent's memories are prisoners. If you build an AI agent today using Mem0, your memories are locked in Mem0. Switch to Zep? You lose everything. Move to a new framework? Start from zero. This is exactly the problem email had in 1970. Every system had its own format. You couldn't send an email from one system to another. Then SMTP was invented. And email became universal. Today I'm publishing TMX v0.1 — the SMTP of AI agent memory. What is TMX? TMX (Truvem Memory eXchange) is an open, model-agnostic JSON format for storing, exporting, and importing AI agent memories across any platform, framework, or provider. It looks like this: { "tmx_version" : "0.1" , "exported_at" : "2026-06-26T20:00:00Z" , "source" : "truvem" , "agent_id" : "my-agent" , "memories" : [ { "id" : "550e8400-e29b-41d4-a716-446655440000" , "content" : "User prefers dark mode and concise responses" , "created_at" : "2026-06-01T08:30:00Z" , "updated_at" : "2026-06-01T08:30:00Z" , "expires_at" : null , "tags" : [ "preference" , "ui" ], "source_model" : "gpt-4o" , "metadata" : {} } ] } That's it. Plain JSON. Human-readable. Portable. Why this matters Right now, the AI agent ecosystem is exploding. Every week there's a new memory provider, a new framework, a new cloud service. But every one of them uses a proprietary format. This means: Developers are locked to their first choice forever Agent memories can't travel between clouds Switching providers = losing everything your agent learned This is the biggest hidden tax in the agentic AI stack. TMX fixes it with a single open spec that anyone can implement — for free, with no approval needed. The 5 core principles 1. Open — No license required. Implement TMX in any product, commercial or otherwise. 2. Model-agnostic — Works with GPT-4, Claude, Gemini, Mistral, Llama, or any future model. 3. Framework-agnostic — LangChain, CrewAI, Mastra, AutoGen — doesn't matter

NYT shifts OpenAI/Microsoft copyright claims after SCOTUS ruling against Sony.

Everyone is selling AI to small businesses right now. Most of it is hype. But some of it is genuinely useful — and knowing the difference can save you thousands in wasted tooling. I run a small agency in Stuttgart that builds websites and automations for local service businesses: coaches, doctors, beauty studios, consultants. Here's what actually moves the needle for them in 2025. What "AI Automation" Actually Means for Small Businesses Forget the generic pitch. For a local service business, AI automation is useful in exactly three places: Client communication at scale — responding to inquiries 24/7 without hiring a receptionist Reducing admin time — intake forms, follow-ups, reminders, invoicing triggers Content creation — but only as a speed boost, not a replacement for your voice Anything beyond that is usually overkill for a business under 10 employees. The One Automation Every Service Business Should Have Automated follow-up after initial contact. Here's the typical flow without automation: Client fills out contact form You see it 4 hours later You write a reply If you're busy, it takes a day Client has already booked elsewhere With automation: Client fills out form Immediate confirmation email ("Got your message, here's how to book a slot") Link to booking calendar You're notified. If they don't book in 48h, a follow-up email goes out automatically This alone converts 20-40% more inquiries into booked clients. No AI model needed — just a simple workflow in n8n, Make, or Zapier. Where LLMs Actually Help Language models (ChatGPT, Claude, etc.) are genuinely useful for small businesses in these areas: Intake Forms → Personalized Responses A coaching client fills out a detailed intake form. Normally, you'd spend 20 minutes reading it and writing a personalized welcome email. With a simple LLM integration: Intake form submitted Webhook fires to n8n LLM reads the form, generates a personalized summary + welcome You review it in 30 seconds and hit send Same personal

Hello, I'm Shrijith Venkatramana. I'm building git-lrc, an AI code reviewer that runs on every commit. Star Us to help devs discover the project. Do give it a try and share your feedback for improving the product. An AI model writes a paragraph. It sounds fluent. It looks convincing. But how do you know whether it's actually good? This deceptively simple question has occupied researchers for more than two decades. Long before ChatGPT, machine translation researchers faced exactly the same problem. Human evaluation was expensive, inconsistent, and painfully slow. If every new model required thousands of humans to compare translations, research would crawl. That necessity gave rise to BLEU , one of the most influential evaluation metrics in AI history. Years later, as language models became better at paraphrasing and reasoning, BLEU started to show its age. Researchers responded with learned metrics like BLEURT and COMET , which use neural networks to judge language much more like humans do. Interestingly, this mirrors software engineering itself. We first wrote simple unit tests, then integration tests, and today we increasingly rely on sophisticated observability systems. Evaluation metrics for LLMs have undergone a similar evolution. Let's see why. Before BLEU: The Evaluation Bottleneck Imagine you're building Google Translate in 2001. Every time your team improves the model, someone has to read thousands of translated sentences and score them. Suppose a single sentence pair takes only 20 seconds to judge. Evaluating 50,000 sentences would require nearly 280 human-hours . Now imagine dozens of experiments every week. Evaluation—not training—quickly becomes the bottleneck. Researchers at IBM, led by Kishore Papineni , introduced BLEU (Bilingual Evaluation Understudy) in 2002 to automate this process. Their idea was surprisingly simple: If a machine translation resembles what professional translators write, it's probably good. This became one of the most cited papers

“We don’t believe this kind of government access process should become the long-term default,” says OpenAI. “It keeps the best tools from users, developers, enterprises, cyber defenders, and global partners who need them.”

How I built an open-source Slack assistant with persistent semantic memory, powered by any LLM and Mem0's managed memory layer — no vector database required. The problem with Slack bots and memory Most AI Slack bots have the memory of a goldfish. Every conversation starts from scratch. You ask it about your sprint goals, it gives a great answer, then three days later you ask a follow-up and it has no idea what you're talking about. You end up re-explaining context constantly. The commercial solution to this is Claude Tag — a Slack integration that maintains genuine conversational continuity. But it's tied to one provider and not open-source. slacktag-oss is our attempt to replicate that experience: a Slack bot with real, semantic, persistent memory that works with any LLM — including ones running entirely on your laptop. What I built A Python Slack bot with: Socket Mode for local dev (no public URL needed), HTTP-ready for prod LangChain to abstract LLM calls across any OpenAI-compatible endpoint Mem0 managed cloud for semantic memory — no Qdrant, no Pinecone, no infra to run Three memory scopes: per-channel, per-thread, per-DM Built-in !clear and !memory commands A clean, extensible architecture you can fork and build on Architecture Before diving into code, here's the full request lifecycle: ┌─────────────────────────────────────────────────────────────┐ │ Slack │ │ @mention in channel ──┐ │ │ DM to bot ──┼──► Slack Events API │ │ Thread reply ──┘ │ │ └───────────────────────────────────│─────────────────────────┘ │ (Socket Mode / HTTP) ▼ ┌─────────────────────────────────────────────────────────────┐ │ slack-bolt (Python) │ │ bot.py ──► router.py ──► handler.py │ │ │ │ │ ┌───────────────┤ │ │ │ │ │ │ ▼ ▼ │ │ Mem0 Client LangChain │ │ (managed) ChatOpenAI │ └────────────────────────────────────────────────────────────-┘ │ ▼ ┌───────────────────────┐ │ Mem0 Managed Cloud │ │ Vector Embeddings │ │ Entity Extraction │ │ Deduplication │ └───────────────────────┘ The ke

The hire marks OpenAI's latest push into India, expanding offices, partnerships and hiring.

One thing I've noticed after using AI for development over the past year is this: The code it generates is usually correct. The architecture slowly isn't. That doesn't happen because AI writes bad code. It happens because architecture rarely erodes all at once. Imagine a modular application with clear boundaries. The billing module talks to the orders module through its public interface. Authentication is isolated. Notifications are independent. Everything is predictable. Now imagine hundreds of AI-assisted commits over the next few months. One suggestion imports an internal class because it already exists. Another bypasses a service layer because it's shorter. A helper gets copied into another module. A database query is duplicated instead of reused. None of those changes are catastrophic. In fact, every pull request probably gets approved. The application still builds. The tests still pass. Customers never notice. Until one day, making a simple change requires touching five different modules because everything has quietly become connected. That's architecture debt. And unlike a failing test, it doesn't show up immediately. One thing I've realized is that our current tooling doesn't really watch for this. Unit tests verify behavior. Integration tests verify interactions. Linters enforce style. Static analysis finds bugs. All of those are important. But none of them are asking questions like: Should this module depend on that one? Did someone bypass a defined boundary? Are we introducing new architectural coupling? Is the overall architecture getting healthier or worse over time? Those questions usually get answered during code review. Or worse, during a production incident. The interesting part is that AI isn't really the problem. If anything, it's doing exactly what we ask it to do. It optimizes for solving the problem in front of it. Architecture, on the other hand, is about protecting the system as a whole. Those are different goals. As AI makes us write code fa

J'utilise la dictée vocale tous les jours depuis six mois. Pas pour taper moins vite. Pour penser plus vite quand je vibe-code avec Claude Code et Cursor. Pis j'ai fini par construire mon propre outil parce que les outils existants me tapaient sur les nerfs d'une façon très précise. Le problème réel Quand tu travailles en tech au Québec, tes phrases ressemblent à ça : "OK fa que je fais un useState pour le component pis je passe le handler en props" Ça, c'est une phrase normale. Personne en tech QC ne parle autrement. Pas parce qu'on est négligents avec la langue. Parce que le vocabulaire technique vient de l'anglais et qu'on le soude naturellement au français au fil de la pensée. Ça s'appelle le code-switching. Et c'est là que la plupart des outils de dictée craquent. Ce que les outils mainstream font mal Dragon NaturallySpeaking Dragon, c'est le vieux standard. Médical, juridique, corporate. Ça coûte environ 500$ en une shot. C'est lourd à installer et à entraîner. Et sa gestion du français québécois avec des termes tech intercalés... c'est en gros zéro. "useState" devient "usé état". "Fa que" devient "faque" parfois, "fake" d'autres fois. C'est aléatoire. T'as intérêt à corriger après chaque phrase. Wispr Flow Wispr Flow est plus moderne. UX propre, cross-platform, et leur gestion du français s'est améliorée. Leur plan Pro coûte 15$/mois, soit environ 144$/an. Mais il y a un problème structurel que leur propre doc admet : la détection de langue se fait par session, pas par mot. Autrement dit : Wispr détecte la langue une fois au début de la session. Si tu commences en français, il reste en mode français jusqu'à la fin. Les mots anglais qui arrivent dans la phrase, il tente de les translittérer en français. "Handler" peut devenir "andler" ou "ender", "props" survit parfois, parfois pas. C'est variable. Pour une phrase de temps en temps avec un mot anglais, ça passe. Pour un vibe-coder québécois qui switch constamment dans la même phrase, ça ne passe pas. Pourquoi

Nvidia has dominated the AI chip market for years, but the era of total dependence might be ending. OpenAI just shared its plans to spice things up with Jalapeño, its custom inference chip built with Broadcom, joining Google, Apple, and SpaceX in a growing list of companies building their way out of single-supplier risk. The goal is less of a […]

The White House asked OpenAI to delay the rollout of its GPT-5.6 AI models, two weeks after Anthropic had to take its most advanced AI models offline.

Less than 24 hours after news broke that OpenAI would stagger its next model release at the request of the Trump administration, that model, GPT-5.6, is here. On Friday, the company unveiled the limited preview of its new GPT 5.6 model suite: Sol, the flagship; Terra, a medium-tier model for "high-volume work"; and Luna, a […]

AI models have progressed to the point where their capabilities have real political consequences. Dealing with those consequences will require collective action.

You don't have a million labeled images or a GPU farm — and you don't need them. Transfer learning lets you stand on a model someone else trained and reach high accuracy with a few examples in minutes. Here's the idea, visualized. ♻️ Race scratch vs transfer: https://dev48v.infy.uk/dl/day17-transfer-learning.html The insight The early layers of a trained network learn general features — edges, textures, shapes — that are useful for almost any vision task. Only the last layers are task-specific. So why relearn edges from scratch? Two ways to do it Feature extraction: freeze the pretrained backbone, replace the final classifier with a small new "head," and train only the head on your data. Fast, needs little data. Fine-tuning: also unfreeze the top few backbone layers and train them at a low learning rate so you adapt without wrecking what they learned. The demo races two accuracy curves: "from scratch" crawls up and plateaus low (not enough data); "transfer learning" starts high and climbs fast. Tweak the example count and freeze/fine-tune to see them respond. Why it matters now This is exactly why fine-tuning an open LLM works: a foundation model already learned language; you adapt it cheaply. Transfer learning is what makes deep learning practical for the rest of us. 🔨 Full recipe (load pretrained → freeze → new head → train → optionally fine-tune low-LR) on the page: https://dev48v.infy.uk/dl/day17-transfer-learning.html Part of DeepLearningFromZero. 🌐 https://dev48v.infy.uk

Hey everyone 👋 Is anyone here playing RF Online Next? I recently built a fan guide website for it: 👉 https://rf-online-next.net RF Online Next Guide — Starter Finder & Beginner Tips New to RF Online Next? Answer 3 questions to get your starter Biosuit, faction lean, and first-day checklist — personalized for your playstyle. rfonlinenextguide.com The idea is pretty simple. When a new MMO launches, information is usually all over the place — Discord messages, random posts, outdated guides, fake code pages, and long videos when you only need one quick answer. So I wanted to make a cleaner guide hub for players who just want to know: how to download and play which faction to pick what Biosuits/classes are good whether there are any real codes how to fix server full/login issues how Mining War / Chip War works what Sacred Weapons do The site focuses a lot on Mining War, the big 450-player faction war between Bellato, Cora, and Accretia. I also tried to keep the content honest. For example, the codes page doesn’t list fake “working codes” just for clicks. If there are no confirmed codes, it says that clearly. From the dev side, I structured the site around search intent instead of a normal blog feed. So the homepage points players directly to the guide they probably need. It also has multilingual sections for different regions, since RF Online Next has players from many countries. Would love to hear feedback from other devs, especially on: site structure SEO approach guide layout content clarity anything that feels confusing If you’re into MMOs, gaming websites, or niche SEO projects, feel free to check it out: 👉 https://rf-online-next.net RF Online Next Guide — Starter Finder & Beginner Tips New to RF Online Next? Answer 3 questions to get your starter Biosuit, faction lean, and first-day checklist — personalized for your playstyle. rfonlinenextguide.com

Check how on-premises AI solutions empower enterprises to safeguard sensitive code, ensure data residency, and maintain full compliance without compromising performance. Why privacy and security matter in AI-powered development? As enterprises increasingly adopt AI to automate code reviews, testing, and vulnerability scanning, ensuring data privacy becomes paramount. Cloud-based AI tools may expose sensitive source code, customer data, or intellectual property to external risks. By contrast, on-premise AI tools allow organizations to keep data within their controlled environments by aligning with data sovereignty and compliance requirements like GDPR and CCPA. According to Gartner, by 2026, 75% of organizations will demand AI solutions that guarantee strong data residency and compliance assurances. What are on-premise AI tools for software development On-premise AI tools are artificial intelligence solutions that are deployed and operated within an organization’s own infrastructure, rather than relying on external cloud services. In the context of software development, on-premise AI allows teams to leverage advanced AI capabilities such as code analysis, automated testing, and security scanning while keeping all data and processes within their own controlled environment. Core components of on-premise AI infrastructure include: Hardware: servers, GPUs, and storage devices physically located on-site or in a private data center. Software: AI models, orchestration tools, and management platforms installed and maintained by the organization. Security Measures: firewalls, access controls, and monitoring systems tailored to the organization’s specific needs. Examples of on-premise AI tools in software development: AI-powered code review platforms installed on internal servers automated vulnerability scanners running within the company’s network machine learning models for test automation, hosted locally. Primary connection to data privacy: on-premise AI ensures that sensit

This is just an opinion of what I experience and am witnessing, but looking at how LLMs scale feels like I've seen it before: with CPUs trying to outrun Moore's Law and break the rules of physics. Heat, power leakage, and diminishing returns made it increasingly expensive to squeeze out even small gains in clock speed. The GHz race shifted because it had to. For LLMs, more compute, more data, more parameters, and everything just keeps getting better? That curve seems to hit a ceiling and innovation needs to succeed the scaling race now. History does not repeat itself, but it rhymes. What learnings can we make from history to "predict" a potential future? History In the early 2000s, CPUs ran into a wall, a very physical one ^^ So makers adapted. Instead of crunching every single watt out of a single core, multi-cores became common. Athlon 64 x2, Pentium D, PS3 with its heavy Cell approach. From linear to parallel. From sequential to multi-threaded (and funny race conditions ;). Talks of distributed systems, SIMD/MIMD and new benchmarking spawned into what we have today. We still use CPUs, but differently. We still have Memory, but think about Cache, RAM, GPU or Unified. Same same, but different. Innovation because of limitation. Present I feel something similar is about to happen to gen AI. Yes, there are improvements in different areas, some in scaling, some optimisation, some performance, but the slope is becoming slippery. The last 12 months went from "Opus 4.5 is the pinnacle" to "What the hell is wrong with Claude?". The perfect (business) storm of scaling execution! But the low-hanging fruits have been eaten and the crops don't grow as fast anymore. Costs rise quickly, latency becomes a constraint, and even large context windows feel more like extensions than breakthroughs. What remains is more incremental, more expensive, and more complex. You could argue the whole venture of "agents" is the same multi-core experience repeating itself. A different kind of orch

I have a confession. Somewhere around day nine of this experiment, I almost quit and went back to my old setup. Not because ChatGPT was bad. Because I was bad at using it. I kept typing half-questions the way I'd type into Google, hitting enter, and getting answers that were technically correct and completely useless. It took me about a week to realize the problem wasn't the tool. It was twelve years of muscle memory. This post is the long version of what happened when I tried to go a full month without my usual stack of developer crutches — Google, Stack Overflow, Regex101, JSONLint, a SQL formatter site, a commit message generator, a pile of bookmarked Docker cheat sheets, and a few other tabs I didn't even realize I kept open until they were gone — and replaced all of it with a single ChatGPT window. I work as a backend-leaning full stack engineer at a small e-commerce company. Python and Django on the server, a chunk of Node for a couple of internal services, Postgres, Docker, and an AWS setup that I inherited rather than designed. Nothing exotic. Which is actually why I think this experiment is useful — most of you reading this aren't working on some bleeding-edge ML pipeline either. You're maintaining stuff, fixing stuff, shipping features under deadlines that someone in another department picked without asking you. So here's what happened. All of it. The good parts, the embarrassing parts, and the parts where I quietly reopened Stack Overflow in an incognito tab because I didn't want my browser history to judge me. TL;DR I tried to replace 12 daily developer tools with ChatGPT for 30 days straight, tracking what worked and what didn't. Google search volume dropped by roughly 70%, but it never hit zero — and I don't think it should. Stack Overflow was the hardest habit to break, and also the one I missed least once I'd broken it. The small utility sites (Regex101, JSONLint, SQL formatters) were the easiest wins. ChatGPT replaced almost all of them outright. Do

Making of Aantraa aantraa.site — AI audio & video translation, caption generator, and viral shorts cutter. Under the Hood I run a small YouTube channel. I'm not a full-time content creator, but YouTube is a solid platform to gain traffic for your online work, business, project, or idea. Aantraa is what I built in a week. The main concept is simple: Video translation into multiple languages Audio translation — including text-to-audio, with MP3 output for Premiere Pro Long-form to shorts — convert YouTube long-form video into short clips At that time, only three features were needed, so website development wasn't the heavy lift. The real work was building APIs, backend infrastructure to integrate AI into video, and dealing with heavy storage. Breaking the execution into steps: How I made Aantraa AI LLM layering and provider Aantraa is heavily dependent on AI APIs — we need reliable infrastructure for LLM providers. OpenRouter, Portkey, Vercel AI SDK labs, and individual APIs for Anthropic, Deepseek, and OpenAI are solid options. I prefer OpenRouter for Aantraa for one reason: multiple model support — it's easy to pick the cheapest capable model for each job. Easy to integrate, strong community support, free model access, and more. AI LLM APIs are needed at almost every stage in the backend: Understanding video context and creating a script Translating the script into target languages Recording the script into MP3 or WAV format Summarising the video Generating captions Cutting videos into shorts Building APIs and servers Each layer needs heavy AI context and prompt engineering. Loop engineering is the trend here — and it's required for aantraa. For example, video translation works in multiple connected steps: Video translation API breakdown AI understands the video, fed into the LLM via the ffmpeg module AI generates a script/caption from the video AI translates the script into the desired language AI generates audio (MP3 or WAV) of the new translation AI glues audio a