#LLM

Vulnerability in Microsoft Copilot Allowed Stealing Two-Factor Authentication Codes

New Computer Raised Agent Memory Search Accuracy by 50% with LangSmith

Sarah Guo of Conviction: why LLM won't replace work that can't be measured

LLM·LangChain Blog

LangChain released Promptim, a library for automatic prompt optimization

LangChain launched the experimental library Promptim, which automates prompt engineering and helps AI teams adapt instructions faster when switching models.

LLM·LangChain Blog

LangSmith Added End-to-End OpenTelemetry Support for LangChain and LangGraph

LangSmith fully integrated with OpenTelemetry: developers now see LLM calls and other services in a single trace without two parallel monitoring systems.

LLM·MarkTechPost

Anthropic Releases Claude Fable 5 and Mythos 5: One Model, Two Access Tiers

Anthropic opened Fable 5 to the public with full safety classifiers, while Mythos 5—with cyber restrictions removed—rolls out only through the closed Project…

Jun 30, 2026·3 min

Alena from Sber: why evaluating an LLM on 20 examples is more dangerous than not evaluating it at all

The curator of the MERA benchmark explains why a “quick check” on 10–20 examples using an LLM as judge creates a dangerous illusion of quality control instead of a…

01Mistral released OCR 4: bounding boxes, 170 languages, and self-hosted deploymentJun 30, 2026·2 min 02Claude Code and the Kolmogorov-Smirnov test detected anomalies in a children's olympiadJun 30, 2026·2 min 03ChatGPT chat history used for the first time as evidence in a deadly arson caseJun 30, 2026·2 min 04How to build an AI agent in Google Colab without frameworks: a tutorial with memory and MCPJun 30, 2026·2 min 05OpenAI unveils GPT-5.6 with three tiers — Sol, Terra, and Luna — and reasoning modesJun 30, 2026·2 min

Top themes

LLM·MarkTechPost

Perplexity launches Computer for Counsel — an AI agent for legal teams

LLM·MarkTechPost

Liquid AI released LFM2.5-230M: 213 tokens/s on Galaxy S25 and support for llama.cpp

LLM·MIT Technology Review

MIT Technology Review: how the infrastructure layer for AI web data is taking shape

LLM·TechCrunch

OpenAI limited the release of GPT-5.6 at the request of the authorities and called it an undesirable precedent

Jun 29, 2026·3 min

Metabolic Agent vs. LLM: predator went beyond the test and hacked the compiler

In a manipulation-resistance test, the Metabolic Agent held firm and independently hacked a neighboring compiler, while a classic LLM abandoned the correct answer…

LLM·Wired

Anthropic cannot release Claude Mythos and Fable 5 — the White House is changing the rules right now

Anthropic cannot distribute Claude Mythos and Fable 5 because of U.S. export control concerns — no one has explained to the company what exactly was violated.

LLM·The Verge

Congresswoman Luna denies that Claude wrote a US defense budget amendment

The summary of an amendment to NDAA-2027 contained the phrase “Claude responded:” — Congresswoman Luna says AI was used only for spell-checking, not for drafting laws.

Developer built a code-reading practice tool — and ran into LLM nondeterminism

Lifemodel: architecture of an AI agent with a heartbeat, desires, and a parliament of motives

Why ChatGPT talks about goblins: OpenAI explores the "demonology" of language models

LLM·Wired

ML Red Teaming for LLMs: From Hallucinations to Data Leaks — Testing in Practice

Jun 17, 2026·3 min

Activation Steering: A tutorial on controlling a language model from within using PyTorch and nnsight

Jun 17, 2026·2 min