Full archive

Launch: Latest AI News

Search reporting on AI markets, models, products, companies, policy, and culture.

Showing 1-13 of 13Results matching "Agent harness"

Yesterday

1 story

YesterdayTools

Launch Tools

Google Launches Fairwind to Give Selected Defenders AI Tools for Finding and Fixing Flaws

The program draws a line between broadly available code-security tooling and restricted access to Gemini 3.8 Flash Cyber, reflecting the dual-use risk of AI that can identify and repair software weaknesses.

3 min

Earlier this week

2 stories

YesterdayProducts

Launch Products

CrowdStrike’s SafeMind Pits Red and Blue AI Agents in an NVIDIA Infrastructure Twin

The new Falcon system combines CrowdStrike’s security data and agent harnesses with NVIDIA’s open Nemotron models. Early tests favor the optimized open pipeline on detection quality, but cover only one attack scenario family.

3 min

Aug 31, 2026Products

Launch Products

Modellix Adds 28+ Models, Including $0 Options, to DeepSeek Harness Beta

The beta gives users of the rapidly adopted coding agent one model-selection layer and web-tool routing, while its free-call pitch still requires a Modellix account and funded balance to create an API key.

3 min

Last week

5 stories

Aug 27, 2026Products

Launch Products

Harness Launches Agent-Ready Repository and AI Review for High-Volume Code Changes

The products pair high-volume repository operations with inherited permissions and merge-time checks, but Harness still places the final decision to ship with people.

2 min

Aug 26, 2026Infrastructure

Launch Infrastructure

AWS Publishes Two AgentCore Paths to Query Cross-Account Knowledge Bases

The useful constraint is not simply access across AWS accounts: generated answers require an assumed role, while many workloads may not need an agent at all.

3 min

Aug 25, 2026Models

Launch Models

DeepSeek V4 Pro Lands on Fireworks With $2.50 CyberGym Solves, but Kimi K3 Scores Higher

The provider’s results frame security-agent selection as a three-way choice among raw solve rate, cost per completed task, and whether a model will reliably execute the required tool workflow.

3 min

Aug 25, 2026Products

Launch Products

Perplexity’s Portable Computer Runs Agents Locally—but Needs a 24GB Nvidia GPU

The product turns a difficult DIY local-AI setup into one application, but its hardware floor and Linux-only launch keep it aimed at well-equipped paid users.

3 min

Aug 24, 2026Products

Launch Products

OpenAI’s $20 ChatGPT Work Brings Codex Agents to Office Apps, With Access Still a Barrier

The product’s promise is a ChatGPT that acts across workplace systems rather than merely answering prompts. Its immediate test is whether non-engineers will grant access and navigate controls built for more autonomous software.

3 min

August 2026

5 stories

Aug 23, 2026Products

Launch Products

Perplexity’s Agent API Adds Managed Connectors for GitHub Code and PR Workflows

The preview feature moves service setup out of individual requests; paired with Sandbox, GitHub access can reach repository edits and pull requests, within the connected account’s permissions.

3 min

Aug 22, 2026Products

Launch Products

Databricks Says Its New Extraction Mode Beats Frontier Models on the Documents That Break Them

Precision Mode is designed for cross-page reasoning, enormous line-item outputs and deeply nested schemas. Databricks reports a seven-point lead over its best tested chunk-and-merge baseline, but the evaluation blends internal and public datasets.

3 min

Aug 22, 2026Models

Launch Models

DeepSeek Gives V4-Flash a Separate Vision API With a 384-Token Image Cap

The experimental model gives developers several familiar ways to submit images and a fixed upper bound on image-token use. It leaves the harder questions—visual accuracy, latency and reliability—for users to test.

3 min

Aug 18, 2026Tools

Launch Tools

TestMu’s New Agent Test Shows What It Can’t Verify

Agent Assurance is built around observable effects rather than an agent’s own account of its work. Its central unresolved test is whether that evidence-based scorecard predicts safer behavior in real systems.

3 min

Aug 16, 2026Tools

Launch Tools

Meta Launches Muse Code With a 92% Discount for Training Rights

The beta agent can modify and check code across large repositories, but its cheapest tier lets Meta train on the prompts and outputs routed through it—a trade teams must assess before plugging in their development work.

5 min