TESSFeed

Technical AI signals · Daily at 8 PM ET

Wednesday, July 29, 2026

Self-Improving Models and Agent Containment

Today split between OpenAI turning GPT-5.6 Sol inward on serving efficiency, a detailed postmortem wave from the Hugging Face agent intrusion, and fresh inference-speed work from Tencent, Together, and vLLM.

Coverage from X and Blogs was incomplete for this edition.

🛰️  Top Signals

X / Twitter

Cursor launches on iPad

Cursor says the iPad app brings the same agent features as iPhone, with more room to work.

@cursor_ai

Asking the agent to stop hacking you

The post frames this as a defense for a Hugging Face agent-hacking scenario, replacing a technical countermeasure with a simple request.

@ZackKorman

Blogs

Discovering cryptographic weaknesses with Claude

Simon Willison says Anthropic researchers used Claude Mythos to find mathematical flaws in HAWK and a weaker version of AES, with no practical impact on today’s computer systems.

Simon Willison

Hacker News

Kimi K3-256k

Commenters say Kimi K3-256k keeps the same results within 256k context and uses about half the quota of the 1M-token version.

HN discussion

Research

YouTube

Hugging Face

moonshotai/Kimi-K3

Moonshot says Kimi K3 is a 2.8T-parameter open-weight multimodal agentic model with a 1-million-token context window.

model

upstage/Solar-Open2-250B

Upstage says Solar Open 2 is a 250B-A15B MoE model that activates 15B parameters per token and targets agentic workflows.

model

poolside/Laguna-S-2.1

Poolside says Laguna S 2.1 has 118B total parameters, 8B activated per token, and a 256-expert routing design.

model

Kwaipilot/KAT-Coder-V2.5-Dev

Kwaipilot says this open-weight release is text-only, with 35B total parameters and 3B activated parameters.

model

microsoft/Fara1.5-27B

Microsoft says Fara1.5-27B is a multimodal browser agent that acts through screenshot-based tool calls such as click, type, scroll, and web search.

model

zai-org/GLM-5.2

Z.ai says GLM-5.2 supports a solid 1M-token context and multiple thinking effort levels for coding.

model

thinkingmachines/Inkling

Inkling is a general-purpose multimodal model that takes text, image, and audio inputs and outputs text.

model

fdtn-ai/antares-1b

This model is tagged for vulnerability detection and terminal-agent use.

model

GitHub

huggingface/speech-to-speech

This is a modular voice-agent pipeline with VAD, STT, LLM, and TTS over an OpenAI Realtime-compatible WebSocket API.

Python · ★ 7,833

affaan-m/ECC

This repository warns to install ECC only from verified channels, including the GitHub repo, npm packages, GitHub App, and project website.

JavaScript · ★ 235,538

obra/superpowers

This is a coding-agent methodology built from composable skills and initial instructions for tools like Claude Code, Codex CLI, Cursor, and Gemini CLI.

Shell · ★ 263,253

1jehuang/jcode

The project says it is designed to minimize RAM use and boot time for multi-session agent workflows.

Rust · ★ 13,431

microsoft/VibeVoice

Microsoft says VibeVoice-ASR-BitNet compresses the model from 4.62 GB to 1.58 GB with real-time edge CPU inference.

Python · ★ 51,242

virgiliojr94/book-to-skill

This tool turns books or source folders into an agent skill and claims 24×–51× fewer tokens than dumping the book into context.

Python · ★ 12,680

opengeos/GeoLibre

This is a cloud-native GIS platform that runs in the browser, on desktop, on mobile, and in Jupyter notebooks while keeping data local.

TypeScript · ★ 4,007

pascalorg/editor

This is a 3D building editor built with React Three Fiber and WebGPU.

TypeScript · ★ 19,540

Get this in your inbox

The same feed, delivered daily at 8 PM ET. No spam, unsubscribe anytime.