标签：#ai

There's a keyboard shortcut to open it up, just like the Mac version.

"Bankruptcy cannot become the new land grab for AI.”

Amazon is expanding shopping integrations across Prime Video, letting viewers discover products tied to thousands of shows, movies, and live sports through X-Ray, its shopping app, and a new Lens-powered “Shop the Scene” feature.

Come inside the mind of a bot trying to convince the internet it's human.

OpenAI has released GPT-6 Astra, a new model focused on coding, computer use, long-running agentic tasks, and cybersecurity, with availability across ChatGPT, Codex, and the OpenAI API. By Daniel Dominguez

Pocket FM uses AI to produce 99% of its new content, helping make content production about 80 times cheaper.

The unnamed software will be built around licensed music from UMG artists who want to participate.

César de la Fuente’s lab uses Codex and ChatGPT to search living and extinct genomes for antimicrobial candidates to fight drug-resistant infections.

Penjual online di Indonesia menerima pesan penipuan setiap hari. Modus yang paling sering muncul ada dua. Pertama, phishing yang mengatasnamakan marketplace, lengkap dengan ancaman akun diblokir dan permintaan kode OTP. Kedua, pembeli gadungan yang mengirim screenshot bukti transfer lalu mendesak penjual mengirim barang sebelum uang masuk. Korbannya hampir selalu sama: pelaku UMKM yang panik dan tidak sempat berpikir. SAPA lahir dari masalah itu. Nama panjangnya Scam Analysis and Protection Assistant. Ini AI agent yang dibangun di atas framework Hermes. Pengguna menempel pesan mencurigakan, SAPA mengembalikan skor risiko, daftar indikator yang terdeteksi, dan langkah aman yang spesifik. Skor deterministik, bukan tebakan model Keputusan pertama yang kami ambil: penilaian risiko tidak boleh bergantung pada LLM. Alasannya praktis. Panggilan model itu lambat, biayanya tidak pasti, dan hasilnya bisa berubah untuk input yang sama. Untuk kasus penipuan, ketidakpastian seperti itu berbahaya. Skor 15 hari ini harus tetap 15 minggu depan. Analyzer SAPA ( analyze.py ) bekerja dengan pencocokan pola dan bobot tetap. Ia membaca pesan, memeriksa lima indikator, menjumlahkan bobotnya, lalu menetapkan level risiko. Tidak ada panggilan API. Tidak ada model yang di-load. Seluruh proses selesai di bawah dua detik di VPS kecil. Lima indikatornya: Indikator Bobot Contoh yang ditangkap Permintaan kredensial (OTP, password, PIN) 5 "kirim OTP untuk membuka akun" Tekanan waktu atau ancaman 2 "akun diblokir dalam 10 menit" Domain mencurigakan 3 TLD .info, .xyz, dan sejenisnya Domain mismatch 5 Link bukan domain resmi brand yang diklaim Manipulasi urgensi 2 Desakan emosional agar bertindak cepat Contoh nyata. Pesan "Akun Tokopedia Anda akan diblokir dalam 10 menit. Klik https://tokopedia-verifikasi.info dan kirim OTP untuk membuka akun." memicu empat indikator sekaligus: permintaan OTP (5), ancaman blokir (2), TLD .info (3), dan domain mismatch karena domain resmi Tokopedia adalah tokopedia.c

I recently built GPT Image 2.5, a web-based AI image generation and editing workspace: https://gpt-image-2-5.org/ Most AI image tools follow the same pattern: choose a model, write a prompt, select a size, click generate, and wait for the result. That workflow is still useful, so GPT Image 2.5 supports the classic image generation flow. But I also wanted to explore a more natural way to create images: generating through conversation. Two ways to create images GPT Image 2.5 currently supports both workflows: Traditional prompt-based image generation Conversational image generation with an AI agent The traditional mode is good when you already know what you want. You can write a prompt, upload a reference image, choose the model and output size, then generate or edit the image directly. The AI agent mode is different. Instead of filling out everything manually, you can chat with the agent about what you want to create. For example, you can describe a product photo, a poster idea, a portrait style, or a marketing visual, and the agent helps turn that conversation into an image generation task. This makes the workflow feel closer to working with a creative assistant: You can start with a rough idea Ask the agent to refine the direction Adjust the style through conversation Explore multiple visual concepts Generate images without needing to write a perfect prompt from the beginning Supported models and use cases The workspace includes Flare, Sunburst, and GPT Image 2. Flare is useful for fast creative exploration. Sunburst is better suited for more detail-focused editing. GPT Image 2 remains available for workflows built around the previous generation. Some example use cases: Product visuals Portraits Posters Campaign concepts Social media images Reference-based image editing Creative visual experiments Why add an AI agent? A lot of people do not naturally think in “prompt format.” They may know the result they want, but not how to describe it in a way that works well fo

Universal Music Group is launching a new AI-powered platform that will allow users to draw from its catalog of licensed music to create song remixes, mashups, and new takes on tracks, according to an announcement on Thursday. The record label is developing the platform through a multi-year licensing agreement with ElevenLabs, a company that specializes […]

When people see ChatGPT or Claude answer fluently in Swahili or Thai, the natural impression is that the model "knows" all languages equally well. It doesn't — and the gap isn't cosmetic. It runs straight through the architecture of how these models read and process text. What follows is a breakdown of where that gap actually comes from, how large it is, and what's being done about it. 1. The problem starts before the model "thinks" — at tokenization An LLM never reads text directly. Text is first cut into pieces (tokens) by a tokenizer, and only those tokens are what the model actually processes. Tokenizers are trained on corpora where English historically dominates — in Llama 3's training data, for instance, English and code made up 95% of the dataset, while every other language combined accounted for just 5%. Because of this skew, the tokenizer ends up "efficient" for English and noticeably less efficient for everything else: models generate text one token at a time, and for languages poorly represented in the tokenizer's vocabulary, this makes generation slower and more computationally expensive. How big is the gap in practice? One study covering 25 European languages and ten models put a concrete number on it: tokenization "fertility" ranges from roughly 1.2 tokens per word for English to roughly 3.1 for Greek and Maltese — a penalty of about 2.5x. Notably, the ranking of "efficient" vs. "inefficient" languages barely changes regardless of register — formal or casual text — with a correlation above 0.97 across styles. A similar picture shows up for languages using a different script. A comparison of several tokenizers on equivalent Polish and English text found that where Polish required 547 tokens, English needed only 377 on the same model (Mistral Small 3.2) — and this gap repeats across most tested models. 2. Why this isn't just about speed — it's about money and quality Tokenization inefficiency isn't only "responses in non-English are slower." Three effect

The previous post made a claim I still stand behind: this site has no server, no database and nothing to patch. It is a repository that becomes HTML. Then there is a button in the bottom right corner that answers questions about me, in full sentences, out of a language model. Both things are true at the same time. The second one is only free at the point where you happen to be looking. Where the agent actually runs Not here. It is a Docker image on a Hugging Face Space , on the free CPU tier, two shared cores. A GitHub Action pushes the repository to the Space on every commit to main , so shipping the agent is still git push , the same gesture that ships the site. The site's half of the contract is one line: export const CHAT_ENDPOINT = ' https://martinimarcello00-personal-cv-langgraph.hf.space/chat ' ; That is the whole dynamic part of a static site. Everything in dist/ is still a file, and the one thing that thinks lives somewhere else behind a URL. The architecture did not get simpler. It moved. Free means it falls asleep A free Space sleeps after 48 hours without traffic, and only paid hardware can be told not to. The next visitor restarts it, which sounds harmless until you read what the container does when it boots: CMD ["/bin/bash", "-c", "python build_rag.py && uvicorn api:app --host 0.0.0.0 --port 7860"] The vector index is not baked into the image. It is rebuilt at startup, every document re-embedded with bge-small-en-v1.5 on those two shared cores before the first request gets an answer. Somebody arriving after a quiet weekend is not waiting for a container to start. They are waiting for an embedding job to finish. So the Space is never allowed to be idle. The monitor that already watches the containers on my home server watches this one too. Uptime Kuma — The thing that notices before I do. Scheduled checks against the handful of endpoints that are supposed to answer, with a notification when one stops. The job I did not expect to give it is keeping some

Excited to begin sharing my journey as a full-stack developer focused on healthcare platform development. I’m currently building my skills with Python, Django, FastAPI, React, JavaScript, SQL, and cloud technologies while exploring secure APIs, healthcare data privacy, AI-powered tools, and scalable patient-management systems. My goal is to create reliable, user-friendly software that improves healthcare operations and patient experiences. I’ll be sharing projects, lessons learned, technical insights, and progress along the way. Open to connecting with developers, healthcare technology professionals, and companies building the future of digital healthcare.

Software engineering has entered a new phase — and it’s no exaggeration to say the community is on fire. The rise of LLMs has dramatically increased productivity and unlocked possibilities that, until recently, felt out of reach. The emergence of highly scalable products built by extremely lean teams — even a single founder powered by AI — makes it clear: this isn’t a passing trend, it’s a structural shift in how software is designed and delivered. Tools like Claude and other AI assistants have fundamentally changed the act of programming. Writing code is no longer the primary bottleneck. The real edge now lies in knowing what to build, how to structure it, and how to integrate complex systems in a resilient way. In this context, it’s natural for senior engineers to question their place — but that uncertainty often comes from an incomplete view of what’s actually happening. In practice, experienced professionals are more valuable than ever. The ability to connect dots, design sustainable architectures, ensure production reliability, and operate systems at global scale hasn’t been replaced — it’s been amplified. In industries like banking and fintech, where precision and robustness are non-negotiable, AI doesn’t replace senior engineers — it empowers those who already know how to play the game at a high level.

Meta has launched its new Muse assistant, marking the company's first real foray into AI-powered productivity tools. The company says its AI agent can "take the busywork off your plate" by helping you with online shopping, emails, trip-planning, and more. I decided to try out the new tool and see how well it performed - […]

Marvel's Wolverine captures just how angry its lead character is. The latest PS5 exclusive from Spider-Man developer Insomniac, Wolverine is a straightforward action game that is at its best when you're (literally) slicing through enemies or thrust into a sequence pulled from a blockbuster movie, like climbing the fuselage of a plane as it careens […]

“The vast majority of cases we find are people who are entitled to claim for something, claiming for that thing,” the researcher told TechCrunch.

Jimmy Kimmel will be interviewing Democratic Texas Senate candidate James Talarico "under unusual circumstances," posting the interview directly to YouTube, rather than airing it on TV during Jimmy Kimmel Live. In Wednesday night's episode, Kimmel said this was out of concern for retaliation from the Trump administration's FCC: [Trump's] FCC has threatened me, threatened our […]

Maven Robotics emerged from stealth today with a $100 million Series A and active deployments.