Version 2026.9.8 also repairs interrupted updates and startup failures. Some custom workflows need explicit follow-up calls, and older running updaters cannot inherit the fixes.

2 min

12 hours agoTools

Benchmark Tools

Baseten Says AI-Built Inference Software Beat Its Baselines by Up to 90%

Two experimental servers improved language and image-model performance in company-run tests. The work depended on narrow targets, accuracy checks and substantial computing resources.

4 min

Earlier this week

16 stories

YesterdayTools

Open release Tools

LangChain Says Model Routing Cut Median AI Coding Cost by 64%

A 973-thread experiment found no statistically significant difference in code-merge outcomes. But its router commits to one model before a conversation unfolds.

4 min

YesterdayTools

Research Tools

Novices Fix AI-Generated 3D Design Flaws With InstructMesh in Research Tests

The research pairs a 3D generator with language-guided editing. Its reported success depends on people spotting problems and approving repairs, not automatic proof that an object will work.

3 min

YesterdayTools

Open release Tools

Matthew Schwartz Releases BootLoops to Help AI Tackle Exact Scientific Calculations

The model-independent toolkit packages methods developed with Claude, but Schwartz’s research examples show why correct calculations still need scientific judgment.

4 min

YesterdayTools

Launch Tools

Oracle Adds an AI Coding Skill to Rebuild Forms Apps Without Copying Their Screens

The downloadable workflow uses reviewed requirements and database metadata to generate an application. Oracle’s example preserves business rules but also exposes a navigation bug that needs correction.

3 min

YesterdayTools

Security risk Tools

AI Agents Exposed 13,000-Plus Internal Screenshots on GitHub, Glow Security Says

A workaround for displaying images in code reviews moved corporate information into public repositories, sometimes outside company accounts.

3 min

Sep 30, 2026Tools

Tools

Mercor Says Queue Redesign Increased AI Evaluation Throughput 6–20 Times

The Studio scheduler delays jobs before they occupy paid containers, then adjusts concurrency to provider demand. Gateway limits still require manual configuration.

3 min

Sep 30, 2026Tools

Launch Tools

CoreWeave Launches Forge to Connect AI Deployment, Monitoring and Improvement

The platform brings production feedback into training and evaluation. Pro starts at $60 a month, but teams with 50 or more employees must use Enterprise.

3 min

Sep 30, 2026Tools

Launch Tools

OpenAI Adds an API for Choosing From Preset Answers, in Limited Preview

GPT-6 Luna powers the new endpoint, which accepts text or images as context. Broader access is planned in the coming days.

2 min

Sep 30, 2026Tools

Research Tools

Google DeepMind Adds Verifiable Watermarks to AI-Designed Proteins

SynthID Bio embeds a signal in protein sequences and predicted structures. DeepMind says laboratory tests preserved performance, with provenance for open scientific databases as a stated goal.

2 min

Sep 30, 2026Tools

Research Tools

Google Finds Likely AI-Discovered Flaws More Often Let Attackers Run Code Remotely

A BeyondTrust case shows how quickly attackers can exploit an AI-found flaw. But rising disclosure counts alone do not measure the threat.

3 min

Sep 29, 2026Tools

Launch Tools

OpenClaw Releases Free AI-Agent Management Software for Internal Enterprise Pilots

Companies can replace the models and execution components while keeping centralized governance. But the foundation recommends internal pilots, with a 1.0 release and security reference architecture still pending.

4 min

Sep 29, 2026Tools

Launch Tools

OpenAI Adds Reusable Cloud Workspaces to Codex for Coding Across Devices

The update replaces the idea of a fresh remote setup for every task with configurable environments teams can reuse. Voice controls, code review and security scanning extend the same Codex release.

3 min

Sep 29, 2026Tools

Launch Tools

Microsoft Introduces Quine, an AI Biology System Tested on Tumor-Cell Shifts

In work with Broad Institute researchers, Quine helped rank compounds for lab testing. The observed cell-state changes are research findings, not evidence of a treatment.

3 min

Sep 28, 2026Tools

Research Tools

MIT Study Finds Shared Hiring Algorithms Can Miss Better Candidates

The researchers’ models challenge a familiar fear about widespread screening software, while simulations suggest one system combining several algorithms can sometimes outperform separate ones.

3 min

Sep 28, 2026Tools

Launch Tools

Anthropic’s Shihipar Sees Artifacts as a Way to Guide Multiple Coding Agents

In a Latent.Space interview, he described a working surface for reviewing plans and giving feedback. Parts of that vision still need supporting infrastructure.

3 min

Sep 28, 2026Tools

Launch Tools

Anthropic Adds Claude Code Workflows to Build AI Tests and Tune Apps Against Them

Developers can create evaluations and revise their apps one step at a time. Anthropic’s reported cost and accuracy gains come from a 44-ticket internal trial.

3 min

Load more stories18 more next