Other ai notis sites
AIYou are hereAI models, prices and agents.ainotis.com HikingTrail openings, hut bookings, and route conditions.hiking.notis.newsPhotographyCameras, lenses and firmware.photo.notis.newsMore areas when we can cover them every day.
Good morning
Member since · Your page
Built on mmBERT-small, the Apache-2.0-licensed classifier takes a state, a question and up to 20 options, and returns one of them.
The changelog lists Opus 5.5 at 4 dollars per million input tokens and 20 dollars per million output tokens.
The paper details production infrastructure for agentic training, spanning container, microVM and full-VM backends behind one SDK.
338 stories · 3921 facts checked · 754 sources from 279 publishers · 53 editions since 22 September
The day's stories and numbers, before work. Free.
You get it each morning. Today's went at .
No case followed yet.
Built on mmBERT-small, the Apache-2.0-licensed classifier takes a state, a question and up to 20 options, and returns one of them.
Supersonic Labs released Julia-1, a 144.3-million-parameter open-weights model for classification and binary judgments rather than text generation. Built on JHU CLSPs mmBERT-small, it accepts a context, a question, and 2 to 20 candidate answers, and its FP32 weights occupy 550.5 MiB. The model artifacts are licensed under Apache 2.0, and the checkpoint evaluation, dated September 24, 2026, is the first release in what the company calls the Julia family of decision models.
The changelog lists Opus 5.5 at 4 dollars per million input tokens and 20 dollars per million output tokens.
The Claude Code changelog from Anthropic lists 114 entries under version 2.1.280, dated September 22, 2026. It fixes writes through a symlinked path being judged by their in-tree spelling: the permission prompt now names where the write actually lands, and acceptEdits, allow rules and auto mode no longer approve a write that lands outside the project tree. The same release adds Claude Opus 5.5 as the default Opus model, with a 1M-token context window priced at 4 dollars/20 dollars per million input/output tokens and 20 cents per million tokens for cache reads.
The paper details production infrastructure for agentic training, spanning container, microVM and full-VM backends behind one SDK.
DeepSeek and collaborators published a paper on DeepSeek Elastic Compute (DSec), a sandbox infrastructure built for large-scale agentic training, describing a unified SDK over FnCall, container, microVM and full-VM execution backends plus a distributed filesystem called Fire-Flyer File System (3FS). The paper reports the production system handles about 3 million sandboxes daily, supports more than 380,000 concurrent sandboxes, sustains over 5,000 sandbox creations per second, and that a single production unit spans around 160 nodes. The paper was submitted to arXiv on September 19, 2026.
On the mornings we publish: the three top stories and up to four short ones. Free.
We email you a link to confirm. An issue may include one sponsor, always labelled Sponsored · Advertisement. Our emails count opens and clicks, not who made them. Unsubscribe in one click. What we keep
Put ai notis on your home screen. No app store.
Put ai notis on this computer. No app store.
Tap Share (the square with an arrow), then Add to Home Screen . If you can't see Share, tap (the three dots) first.
If you're a member, you'll sign in once more inside the app.
Models · 27 SEP
Built on mmBERT-small, the Apache-2.0-licensed classifier takes a state, a question and up to 20 options, and returns one of them.
Checked 27 Sep, 04:30 · Hugging Face (Supersonic Labs model card) · 1 sourceTooling · 27 SEP
The changelog lists Opus 5.5 at 4 dollars per million input tokens and 20 dollars per million output tokens.
Checked 27 Sep, 04:43 · Anthropic · 1 sourceResearch · 27 SEP
The paper details production infrastructure for agentic training, spanning container, microVM and full-VM backends behind one SDK.
Checked 27 Sep, 04:30 · arXiv (DeepSeek) · 1 sourceThat's today's news: three stories.
You can join ai notis soon, to follow cases and keep stories on every device.
Follow us on Instagram and Threads
On the mornings we publish: the three top stories and up to four short ones. Free.
We email you a link to confirm. An issue may include one sponsor, always labelled Sponsored · Advertisement. Our emails count opens and clicks, not who made them. Unsubscribe in one click. What we keep
Put ai notis on your home screen. No app store.
Put ai notis on this computer. No app store.
Tap Share (the square with an arrow), then Add to Home Screen . If you can't see Share, tap (the three dots) first.
If you're a member, you'll sign in once more inside the app.
Built on mmBERT-small, the Apache-2.0-licensed classifier takes a state, a question and up to 20 options, and returns one of them.
Supersonic Labs released Julia-1, a 144.3-million-parameter open-weights model for classification and binary judgments rather than text generation. Built on JHU CLSPs mmBERT-small, it accepts a context, a question, and 2 to 20 candidate answers, and its FP32 weights occupy 550.5 MiB. The model artifacts are licensed under Apache 2.0, and the checkpoint evaluation, dated September 24, 2026, is the first release in what the company calls the Julia family of decision models.
Built on mmBERT-small, the Apache-2.0-licensed classifier takes a state, a question and up to 20 options, and returns one of them.
The changelog lists Opus 5.5 at 4 dollars per million input tokens and 20 dollars per million output tokens.
The Claude Code changelog from Anthropic lists 114 entries under version 2.1.280, dated September 22, 2026. It fixes writes through a symlinked path being judged by their in-tree spelling: the permission prompt now names where the write actually lands, and acceptEdits, allow rules and auto mode no longer approve a write that lands outside the project tree. The same release adds Claude Opus 5.5 as the default Opus model, with a 1M-token context window priced at 4 dollars/20 dollars per million input/output tokens and 20 cents per million tokens for cache reads.
The changelog lists Opus 5.5 at 4 dollars per million input tokens and 20 dollars per million output tokens.
The paper details production infrastructure for agentic training, spanning container, microVM and full-VM backends behind one SDK.
DeepSeek and collaborators published a paper on DeepSeek Elastic Compute (DSec), a sandbox infrastructure built for large-scale agentic training, describing a unified SDK over FnCall, container, microVM and full-VM execution backends plus a distributed filesystem called Fire-Flyer File System (3FS). The paper reports the production system handles about 3 million sandboxes daily, supports more than 380,000 concurrent sandboxes, sustains over 5,000 sandbox creations per second, and that a single production unit spans around 160 nodes. The paper was submitted to arXiv on September 19, 2026.
The paper details production infrastructure for agentic training, spanning container, microVM and full-VM backends behind one SDK.
That's today's news: 3 stories.Archive
Built on mmBERT-small, the Apache-2.0-licensed classifier takes a state, a question and up to 20 options, and returns one of them.
Checked 27 Sep, 04:30 · Hugging Face (Supersonic Labs model card) · 1 sourceRead
The changelog lists Opus 5.5 at 4 dollars per million input tokens and 20 dollars per million output tokens.
Checked 27 Sep, 04:43 · Anthropic · 1 sourceRead
The paper details production infrastructure for agentic training, spanning container, microVM and full-VM backends behind one SDK.
Checked 27 Sep, 04:30 · arXiv (DeepSeek) · 1 sourceRead
Built on mmBERT-small, the Apache-2.0-licensed classifier takes a state, a question and up to 20 options, and returns one of them.
The changelog lists Opus 5.5 at 4 dollars per million input tokens and 20 dollars per million output tokens.
27 SEP · Anthropic · 1 sourceThe paper details production infrastructure for agentic training, spanning container, microVM and full-VM backends behind one SDK.
27 SEP · arXiv (DeepSeek) · 1 sourceShort of the target: three today, not five.
Join to keep it on every device. Free, with just your email.
You can join soon.
Prefer email? The morning email brings you the top stories.
We keep your email and what you choose to save. There are no passwords. Our emails count opens and clicks, not who made them. What we keep