ENFR
8news

Tech • IA • Crypto

TodayMy briefingVideosTop articles 24hArchivesFavoritesMy topics

OpenAI Realtime Voice, GPT‑5.5 Instant, Graphify Surge

AIFriday, May 8, 2026· 12 videos

Briefing

0:00 / 0:00

OpenAI unveils GPT Realtime 2

OpenAI introduced GPT Realtime 2, GPT Realtime Translate, and GPT Realtime Whisper, pushing voice into a primary interface for AI. The system enables real-time multilingual translation across ~70 languages, delivering speech output mid-sentence for natural flow. It supports 128,000-token context and near GPT‑5-level reasoning, allowing multi-step actions during live conversations. Benchmarks show accuracy rising to 96.6% on Big Bench Audio, signaling major gains in reliability.

GPT‑5.5 Instant cuts hallucinations

OpenAI launched GPT‑5.5 Instant as its new default general-purpose model, replacing GPT‑5.3 Instant. The model delivers faster responses with ~50% fewer hallucinations, targeting everyday tasks like writing and coding. It emphasizes shorter outputs and reduced token usage, lowering operational costs while maintaining accuracy. OpenAI positions it as suitable for about 95% of use cases, reserving heavier models for deep reasoning.

OpenAI vs Anthropic enterprise push

OpenAI and Anthropic simultaneously launched service-focused divisions—The Deployment Company and Enterprise Services. Both aim to embed AI directly into corporate environments, handling integration, compliance, and long-term operations. The shift reflects a broader move from model development to enterprise adoption and recurring revenue. Competition is intensifying across sectors like finance, healthcare, and insurance, with consulting firms now under pressure.

Graphify explodes to 500K downloads

Graphify surged past 500,000 downloads and 43,000 GitHub stars, highlighting demand for persistent AI memory. The tool builds knowledge graphs from code and documents, enabling structured, reusable context across sessions. Its neuro-symbolic approach links data relationships rather than relying on raw token input. This reduces costs and dramatically improves developer workflows like onboarding and system understanding.

DeepSeek TUI dominates GitHub trending

DeepSeek TUI, powered by DeepSeek V4, rapidly exceeded 10,000 GitHub stars with explosive daily growth. The terminal-native agent can edit files, run commands, manage Git, and coordinate tasks without leaving the CLI. Its tight optimization around a 1M-token context window and low-cost inference drives adoption. The project’s rise underscores growing demand for keyboard-first, agentic coding environments.

Cortex brings AI into Chrome

Cortex launched a Chrome extension that lets AI operate داخل real user sessions with access to tabs, cookies, and logins. This enables automation across full web apps, bypassing limitations of traditional plugins. The system runs tasks in parallel within dedicated tab groups, minimizing workflow disruption. It can execute multi-step actions like research, sentiment analysis, and structured data extraction.

Google doubles down on AI health

Google introduced Fitbit R and expanded its Google Health platform with Gemini-powered coaching. The screenless wearable focuses on continuous tracking, offering 7-day battery life and passive biometric monitoring. AI analyzes metrics like heart rate and sleep to deliver personalized recommendations via subscription. The strategy signals a shift toward predictive, AI-driven healthcare ecosystems built on constant data streams.

RAG workflows replace context overload

Developers are moving away from dumping full documents into models like ChatGPT or Claude, citing “context rot.” Modern pipelines use vector databases and Retrieval-Augmented Generation (RAG) to fetch only relevant data chunks. This improves speed, accuracy, and cost efficiency while avoiding token limits. Tools like Obsidian remain useful for storage but require proper RAG integration to scale effectively.

Videos covered

Previous briefings · AI