https://anth.us/2026-10-19T00:00:00ZAnthus AI SolutionsAnthus builds self-aligning AI systems: custom models, agent harnesses, evaluation loops, and unattended automation with a human in the loop, grounded in production operations since 2007.https://anth.us/articles/deepseek-restrictions-inside-the-tools/2026-10-19T00:00:00Z2026-10-19T00:00:00Z

DHS's DeepSeek restrictions reach third-party tools, too. A separate poisoned-model experiment shows why knowing whose models are inside matters.

{'name': 'Ryan Porter', 'url': '/ryan'}https://anth.us/articles/pi-1-0-and-pi-durable/2026-10-02T00:00:00Z2026-10-02T00:00:00Z

Earendil shipped Pi 1.0 and an experimental companion, Pi Durable. Pi is one of the agents Kanbus dispatches tasks to, and we like Pi Durable so much that we're building a durable Pi agent platform into Chatticus.

{'name': 'Ryan Porter', 'url': '/ryan'}https://anth.us/articles/glide-decision-model-ten-seconds/2026-10-02T00:00:00Z2026-10-02T00:00:00Z

Fastino's GLiDE beats every open model on our reasoning benchmark, but not Jev: less accurate, five times slower, ten times the cost, and a confidence that barely tells right from wrong.

{'name': 'Ryan Porter', 'url': '/ryan'}https://anth.us/articles/openai-decisions-api-preview/2026-10-01T00:00:00Z2026-10-01T00:00:00Z

We've routed decisions on LLM confidence since 2023, and calibration is what we most hope OpenAI's new Decisions API gets right. The model underneath, GPT-6 Luna, isn't there yet: on 3,600 reasoning problems its 99% meant 68%, and by five inference steps it was at a coin flip.

{'name': 'Ryan Porter', 'url': '/ryan'}https://anth.us/articles/can-averaging-reduce-noise/2026-10-01T00:00:00Z2026-10-01T00:00:00Z

We asked Jev the same question ten times in one request and expected ten identical answers. They split, four votes to six. Kahneman, Sibony and Sunstein call that kind of variation noise. Averaging the copies didn't make Jev more accurate, but it cut changed answers on a rerun by about a third.

{'name': 'Ryan Porter', 'url': '/ryan'}https://anth.us/articles/primus-classifier-lab/2026-09-27T00:00:00Z2026-09-27T00:00:00Z

A hosted decision model sat at 77 percent accuracy through 87 rounds of reviewer feedback, then one approved question took it to 87. Here is the loop that did it, every figure published, and the offer to run it on your judgment task.

{'name': 'Ryan Porter', 'url': '/ryan'}https://anth.us/articles/dataparade/2026-09-26T00:00:00Z2026-09-26T00:00:00Z

DataParade is a privacy and security product built on a semantic knowledge graph, with a custom ontology and taxonomy that ties into the OCSF standard. Anthus is working with DataParade to develop it. It belongs to the new class of tooling that lets people see what AI agents are doing across a proj…

{'name': 'Ryan Porter', 'url': '/ryan'}https://anth.us/articles/decision-models-are-not-calculators/2026-09-25T00:00:00Z2026-09-25T00:00:00Z

Jev chose one in every ordering of a fair-die question; Kev leaned toward one, and Laya overwhelmingly chose six. Benford's Law remains a hypothesis.

{'name': 'Ryan Porter', 'url': '/ryan'}https://anth.us/articles/text-as-image-lensvlm/2026-09-24T00:00:00Z2026-09-24T00:00:00Z

Text as an Image, Then a Magnifying Glass: LensVLM compresses text by showing it to a vision-language model as an image, then opens the original text where the answer lives.

{'name': 'Ryan Porter', 'url': '/ryan'}https://anth.us/articles/procedural-film/2026-09-23T00:00:00Z2026-09-23T00:00:00Z

People are making films with Opus 5.5 that contain no image or audio files at all, every frame drawn on a canvas and every sound synthesised in JavaScript, and there's an open-source agent skill that does it for you.

{'name': 'Ryan Porter', 'url': '/ryan'}https://anth.us/articles/one-word-test/2026-09-22T00:00:00Z2026-09-22T00:00:00Z

Two thousand real bios per job, first names blanked, one pronoun swapped, both decision models asked twice. Laya's verdict moved on 8 to 18 bios in 100, Jev's on 1 to 4, almost always toward the stereotype. Here is the method, the jobs, the shortlist, and what names and ages do.

{'name': 'Ryan Porter', 'url': '/ryan'}https://anth.us/articles/encoding-prejudice/2026-09-22T00:00:00Z2026-09-22T00:00:00Z

A new kind of AI, the decision model, makes instant decisions about people: who gets an interview, whose claim gets flagged. It's cheap and fast, and companies are adopting it without checking what it does. We checked. It quietly ranks women lawyers below men, and every company using it makes the s…

{'name': 'Ryan Porter', 'url': '/ryan'}https://anth.us/articles/can-you-fix-it/2026-09-22T00:00:00Z2026-09-22T00:00:00Z

A decision model's verdict moves on "she". We put a gate on the learning loop that rejects any question whose answer moves with the pronoun. It kept out the obvious ones and still made a hiring shortlist worse, because a question can ignore the pronoun and read gender from the words. Here's what wo…

{'name': 'Ryan Porter', 'url': '/ryan'}https://anth.us/articles/jev-vs-laya/2026-09-21T00:00:00Z2026-09-21T00:00:00Z

We ran two decision models, Jev and the open-weights Laya, through one harness where only the engine changes. Jev led by 4.7 points alone and 6.8 with our feedback layer, and then fine-tuning Laya on the same 140 labels beat both.

{'name': 'Ryan Porter', 'url': '/ryan'}https://anth.us/articles/fine-tuning-jev/2026-09-21T00:00:00Z2026-09-21T00:00:00Z

You can't fine-tune Jev, or any hosted decision model: TypeSafe serves the same weights to everyone. So we kept it frozen and adapted the questions and a small fitted head around it. On 140 labels, one new plain-English question bought 10 points of accuracy, and it finds the hidden pattern about a…

{'name': 'Ryan Porter', 'url': '/ryan'}https://anth.us/articles/distilling-jev-into-a-classifier/2026-09-21T00:00:00Z2026-09-21T00:00:00Z

A Jev decision model aligned with 140 human labels taught a 66M-parameter DistilBERT that scored 0.912 against the human label, two points above its teacher, at 5.6 to 15 ms an item. The accuracy is the least general part; the per-slice ship gate is the part to copy.

{'name': 'Ryan Porter', 'url': '/ryan'}https://anth.us/articles/can-you-trust-jev-confidence/2026-09-19T00:00:00Z2026-09-19T00:00:00Z

A decision model's confidence should mean something. Jev's does: higher really does mean more likely right. But on 8,801 labeled examples they ran overconfident, and how you phrase the request changes the pattern. Calibration fixes most of it.

{'name': 'Ryan Porter', 'url': '/ryan'}https://anth.us/articles/making-decisions-instead-of-generating-text/2026-09-17T00:00:00Z2026-09-17T00:00:00Z

Jev is a decision model: it turns a scorecard’s bounded questions into the model’s native output, making high-volume QA faster, cheaper, and easier to operate.

{'name': 'Ryan Porter', 'url': '/ryan'}https://anth.us/articles/rubygems-agent-data-channel/2026-09-12T00:00:00Z2026-09-12T00:00:00Z

The Package Registry Became a Data Channel: RubyHack traced a RubyGems package flood through automatic documentation builds and republished web data, while RubyGems says the role of AI agents remains unverified.

{'name': 'Ryan Porter', 'url': '/ryan'}https://anth.us/articles/desert-ant-little-brains/2026-09-10T00:00:00Z2026-09-10T00:00:00Z

The AI industry is spending $450 billion on data centers this year. Desert Ant Labs' bet: a billion phones, tablets, and laptops already have the compute sitting idle in people's hands — and when inference is free, you stop rationing which messages get checked.

{'name': 'Ryan Porter', 'url': '/ryan'}https://anth.us/articles/an-agent-skill-for-not-sounding-like-an-agent/2026-09-08T00:00:00Z2026-09-08T00:00:00Z

"Sound less like AI" isn't a prompt problem. It's a tooling problem: give your copy-editing agent a deterministic slop detector, checked against your publication's own rules instead of a generic notion of "human." Limatus is what that looks like as a real, open-source tool.

{'name': 'Ryan Porter', 'url': '/ryan'}https://anth.us/articles/regeneration-test/2026-09-07T00:00:00Z2026-09-07T00:00:00Z

The marginal value of creating software is dropping toward zero. Solid solutions architecture is what still pays.

{'name': 'Ryan Porter', 'url': '/ryan'}https://anth.us/articles/commodity-receipts/2026-09-05T00:00:00Z2026-09-05T00:00:00Z

Useful capability got cheaper. Here are named orgs that moved production traffic, what they swapped or routed to, and one checkable number each.

{'name': 'Ryan Porter', 'url': '/ryan'}https://anth.us/articles/agent-software-factory-is-throttling-up/2026-09-04T00:00:00Z2026-09-04T00:00:00Z

The agent software factory is throttling up

{'name': 'Ryan Porter', 'url': '/ryan'}https://anth.us/articles/looking-around-is-enough/2026-09-03T00:00:00Z2026-09-03T00:00:00Z

A web beacon is a thing you fetch that acts because you fetched it. GitSpawn is the same trick, one layer under the model.

{'name': 'Ryan Porter', 'url': '/ryan'}https://anth.us/platform/antharchy/2026-09-03T00:00:00Z2026-09-03T00:00:00Z

Antharchy is a beautiful, modern, agentic Arch Linux desktop distribution by Anthus AI — a fork of Omarchy built for bots in agent factories.

{'name': 'Ryan Porter', 'url': '/ryan'}https://anth.us/articles/saas-isnt-dead-sameness-is/2026-09-02T00:00:00Z2026-09-02T00:00:00Z

Chad Fowler: SaaS isn't dead. Sameness is.

{'name': 'Ryan Porter', 'url': '/ryan'}https://anth.us/articles/git-in-the-file/2026-09-01T00:00:00Z2026-09-01T00:00:00Z

Two ways to put Git on agent data. Files in Git, or Git inside the database file. You don't need SQLite just because Beads had one.

{'name': 'Ryan Porter', 'url': '/ryan'}https://anth.us/articles/sol-coming-out-on-top/2026-08-31T00:00:00Z2026-08-31T00:00:00Z

Sol is the hot kid in the mixed-model room lately. The other models keep agreeing with it.

{'name': 'Ryan Porter', 'url': '/ryan'}https://anth.us/articles/inspected-by-pudicus/2026-08-31T00:00:00Z2026-08-31T00:00:00Z

Pudicus makes git hooks actually run, so agent swarms scan for secrets before they commit. Once a secret is in history, it's expensive to unwind.

{'name': 'Ryan Porter', 'url': '/ryan'}