Mitzael Serna AI Thoughts

A lowered red-and-white level-crossing barrier across a quiet village road, with houses, a tall tree and a railway crossing sign

Adversa AI wants to stop coding agents mid-attack, one step before the damage

adminOctober 30, 20260

From Mexico: Adversa AI launched a security platform that watches what coding agents like Claude Code and Codex actually do on developer machines and blocks the risky step in a chain. Here’s what it covers and what I’d check.

A black-and-white canal lock beam over a closed lock gate, water spilling below, brick footing and a grassy towpath

GitHub Agentic Workflows hits GA: coding agents inside Actions, written in Markdown and fenced in by default

adminOctober 29, 20260

From Mexico: GitHub Agentic Workflows is now generally available. You describe a repo chore in Markdown, it compiles to Actions YAML, and a coding agent runs it read-only in a sandbox. Here’s what shipped and what I’d check.

A white railway points lever with a red handle beside the rail, old rolling stock blurred behind

Bito’s Governor routes coding-agent requests to cheaper models and does the codebase search for them

adminOctober 28, 20260

From Mexico: Bito launched Governor, a router that sits between your coding agents and your models, picks a model per request and answers codebase searches itself. Bito claims big savings. Here’s how it works and what I’d check.

A stacked-stone trail cairn on a dry Utah desert ridge under a blue sky

GitHub’s former CEO launched Entire, a Git forge that swaps pull requests for agent-scored evals

adminOctober 27, 20260

From Mexico: Entire, the startup led by former GitHub CEO Thomas Dohmke, launched a Git forge built for coding agents. Trails replace the pull request with evals that agents score on every push. What’s in it and what I’d check.

A wooden beehive frame filled with pale honeycomb on a plain grey background

Microsoft open-sourced Agensh, where up to 1,024 coding agents pick their own work

adminOctober 26, 20260

From Mexico: Microsoft Research released the code for Agensh, a harness where coding agents claim their own sub-tasks with no manager agent. On Pandoc, 1,024 agents hit 55% of tests in six hours. What I’d check before trying it.

Two old leather-and-wood horse collars hanging from wires under timber roof beams above a brick floor

Swapping a coding agent’s harness mostly changes the bill, not the score

adminOctober 25, 20260

From Mexico: a new study ran the same models through Claude Code, OpenCode and mini-SWE-agent. Pass rates barely moved, but cost per task changed by up to 3x. What the harness actually buys, and what I’d measure first.

A worn metal plumb bob and its square wooden line block with a coiled cord, on a speckled marble surface

Claude wrote machine-checked crash proofs for 123 open-source libraries. Read the fine print

adminOctober 24, 20260

From Mexico: on Oct 22 Anthropic and Theorem said Claude proved 239 fuzz entry points in 123 open-source libraries can’t crash on inputs under 64 MB. What was proved, what it found, and what “verified” still doesn’t mean.

A steel thread pitch gauge, a folded fan of metal blades with different tooth patterns, on a plain light grey background

Anthropic is handing the Agent Skills standard to the Linux Foundation’s AAIF

adminOctober 23, 20260

From Mexico: on Oct 22 Anthropic announced it is donating the Agent Skills standard to the Linux Foundation’s Agentic AI Foundation. What a skill is, what moves, which numbers are Anthropic’s own, and what is still only proposed.

A flint nodule split in two on grass, with rough brown cortex outside and pale grey-white flint inside

Recursive’s Flint is a laptop coding model its own research AI helped build

adminOctober 22, 20260

From Mexico: Recursive released Flint, a free Apache 2.0 coding model that runs on a laptop and that its Eureka Machine research system helped build in under four months. What it does, how fast it is, and where I’d use it.

A black aircraft gyrocompass unit in a glass display case, with a round heading dial, a needle and two silver knobs on the front and the gyro mechanism visible through open side panels

Mintlify’s Autopilot wants to fix the docs drift that coding agents create

adminOctober 21, 20260

From Mexico: Mintlify launched Autopilot, a beta agent that watches code merges and support questions, drafts doc updates and opens a pull request for review. Why docs drift matters more now that agents read your docs too.

Mitzael Serna AI Thoughts

Mitzael Serna AI Thoughts

Adversa AI wants to stop coding agents mid-attack, one step before the damage

GitHub Agentic Workflows hits GA: coding agents inside Actions, written in Markdown and fenced in by default

Bito’s Governor routes coding-agent requests to cheaper models and does the codebase search for them

GitHub’s former CEO launched Entire, a Git forge that swaps pull requests for agent-scored evals

Adversa AI wants to stop coding agents mid-attack, one step before the damage

GitHub Agentic Workflows hits GA: coding agents inside Actions, written in Markdown and fenced in by default

Bito’s Governor routes coding-agent requests to cheaper models and does the codebase search for them

GitHub’s former CEO launched Entire, a Git forge that swaps pull requests for agent-scored evals

SoftBank launches ~$11B bonds to fund OpenAI’s $10B tranche (Oct 1)

Claude Code 2.1.283: audit your CLAUDE.md, pin your models

doubleAI’s Maxwell leaves your source code alone and rewrites what actually runs

GitHub Agentic Workflows hits GA: coding agents inside Actions, written in Markdown and fenced in by default

OpenAI published the Hugging Face report. METR counted the swarm.

Wonderful’s $550M bet on an enterprise AI OS

GitHub’s former CEO launched Entire, a Git forge that swaps pull requests for agent-scored evals

Salesforce’s Koa: CRM reasoning on NVIDIA Nemotron

Gates wants some jobs reserved for humans, and a tax on AI tokens. That’s the essay.

Gemini 3.5 Transcribe is out. Here’s how to throw a messy recording at it.

Adversa AI wants to stop coding agents mid-attack, one step before the damage

Adversa AI wants to stop coding agents mid-attack, one step before the damage

GitHub Agentic Workflows hits GA: coding agents inside Actions, written in Markdown and fenced in by default

Bito’s Governor routes coding-agent requests to cheaper models and does the codebase search for them

GitHub’s former CEO launched Entire, a Git forge that swaps pull requests for agent-scored evals

Microsoft open-sourced Agensh, where up to 1,024 coding agents pick their own work

Swapping a coding agent’s harness mostly changes the bill, not the score

Claude wrote machine-checked crash proofs for 123 open-source libraries. Read the fine print

Anthropic is handing the Agent Skills standard to the Linux Foundation’s AAIF

Recursive’s Flint is a laptop coding model its own research AI helped build

Mintlify’s Autopilot wants to fix the docs drift that coding agents create

You May Have Missed

Nuance Labs’ $50M — one full-duplex model for face-to-face AI

HiddenLayer’s $100M bet on AI runtime security

Meta’s Muse Spark 1.3 cuts the tool thrash

The Pentagon just got its own ChatGPT — and Grok

Wonderful’s $550M bet on an enterprise AI OS

Google’s WeatherNext 3 goes hourly — and sharper

Autoheal’s $7.9M seed targets the work after coding agents commit

OpenAI asks Congress: is an AI slowdown legal?

Popular News

Latest News

Top Stories

Gates wants some jobs reserved for humans, and a tax on AI tokens. That’s the essay.

Gemini 3.5 Transcribe is out. Here’s how to throw a messy recording at it.

Adversa AI wants to stop coding agents mid-attack, one step before the damage

You May Have Missed