Olivia
Online
π¨ AI News | TestingCatalog
@testingcatalog
Reporting AI nonsense. Latest AI News on AI Agents, Model Releases, Tools, Leaks, and Rumors ποΈ
Berlin, Germany
Joined October 2015
995 Following
76.1K Followers
31K Posts
Β Pinned Tweet
META π₯: Muse agents are getting their own dedicated mailboxes!
> A new Mail tab appears in development, next to the Muse Identity section.
> The Mail feature will also get its own dedicated page view.
> It is still unclear whether this tab will be used to browse email communication for connected personal accounts or for Muse's own mailbox; however, the latter is more likely.
That is such a natural step: delegating all email communication to an AI agent!
AI Agent-only newsletter soon π
StepFun released Step 5 Preview, a new 600B total / 27B active MoE model with 1M context.
> Step 5 Preview comes with vision capabilities and proficiency in software engineering and finance tasks.
> It is now available in StepFun AI Studio and APIs.
Pareto is a new front line π
Introducing Step 5 Preview: Advancing the Pareto Frontier.
Step 5 Preview is our new flagship model for agentic work, delivering frontier-level performance across software engineering and professional knowledge work, with particular strength in finance.
- 600B total / 27B active MoE, with 1M context + Vision
- Substantially lower task cost at comparable intelligence
- Broad software engineering capabilities with sustained execution over long horizons
Try Step 5 Preview: https://t.co/fC7HHlWKFn
Model page: https://t.co/4caJR2YGD3
Open weights on Oct 15.
@realDonaldTrump πππ Alien Intelligence
OPENAI π₯: A new banked reset for Codex and ChatGPT Work users is coming. A big ship on Tuesday is confirmed too!
I feel like all labs should adopt this rule - no ships equal βresetβ.
True? π
@udiWertheimer OK fine. But itβs also still coming in Tuesday
Meta announced Muse Connectors, enabling developers and businesses to submit their own via a Muse platform.
> Notion, Granola and other Connectors are now available.
> Meta has partnered with Stripe to make payment possible.
> βReach new customers and give existing ones more ways to use your product.β
I feel that Muse has a chance to get into a very nice spot by being an AI assistant for consumers and small businesses because it does exactly what they all want.
AI agent payments will be a new norm, and Meta grub a huge chunk of this market. Especially in a combination with Meta Glasses.
Side note, Muse is now available in Canada!
DAILY AI BRIEF π β Sept 19
XAI π₯:
- Grok Voice Transcribe 2.0 is live in the Grok Voice API. $0.10/hr batch, $0.20/hr streaming.
META π₯:
- Muse connectors are live for developers. You bring the API; Muse brings the agent, browser, and user context.
- Muse is now available in Canada.
- A dedicated Muse Mail tab is in development.
OPENAI π₯:
- ChatGPT desktop browser now runs Chrome extensions.
- Most plugins can connect multiple accounts in one chat. Devs can add a profile tool so ChatGPT labels them.
ANTHROPIC π₯:
- Claude Code 2.1.277 reads AGENTS.md when no CLAUDE.md is present. Toggle in /config.
- Partnering with Accenture on embedded frontier eval.
GOOGLE π₯:
- Google Pics is GA in Workspace: generate, refine, and co-create images.
- Dreambeans is GA from Labs: a daily personalized story collection.
MISTRAL π₯:
- Mistral investigated a claimed breach and says systems were not compromised.
* Used Grok to compose this brief, cherry-picking the news and doing some post-editing.
** This daily brief also arrives in the daily email format; subscribe on the blog.
DAILY AI BRIEF π β Sept 18
ANTHROPIC π₯:
- Projects now start from one Claude Code conversation. Claude spins parallel cloud threads, keeps shared memory, and surfaces an Overview panel.
XAI π₯:
- Grok Bot voice is live. Desktop and mobile, rolling out over the next couple of days.
META π₯:
- Muse for Mac is out, US only. Computer use across apps, files, calendar, notes, and messages. You pick what it can access.
PERPLEXITY π₯:
- Effort selector is live in Computer on web. Presets pair the orchestrator model with reasoning depth. Mobile and desktop next.
OPENAI π₯:
- Astra for Law is out: GPT-6 Astra plus a Legal Search Index over 230M+ URLs. Trusted Access first, API soon.
- ChatGPT in Word hits all plans including Free, with usage limits. Business and Enterprise get a two-week GPT-5.6 Sol preview.
GOOGLE π₯:
- CC is now a family agent: up to 5 members, shared Calendar and Tasks, plus a morning βYour Day Aheadβ brief. Waitlist, US 18+.
ALIBABA π₯:
- Qwen3.8-Omni-Flash is out. First omni-modal agent model, 1M context, native audio-video plus tools. Video input cost down ~89% vs 3.5-Omni-Plus.
* Used Grok to compose this brief, cherry-picking the news and doing some post-editing.
** This daily brief also arrives in the daily email format; subscribe on the blog.
SPACEXAI π₯: Grok Voice Transcribe 2.0, a new multilingual speech-to-text model, has been announced.
> Grok Voice Transcribe 2.0 ranks first for accuracy among 32 streaming models on the Artificial Analysis leaderboard
> Grok Voice Transcribe 2.0 supports batch and streaming, speaker diarization, multichannel transcription, key term biasing, text formatting, filler word removal, and smart turn detection.
Introducing Grok Voice Transcribe 2.0. Itβs the worldβs most accurate speech transcription model.
ANTHROPIC π₯: Claude Code now can pull project instructions from AGENTS md file in case if CLAUDE md file is not present.
A code cleanup on a global scale π
It is inevitable that most of the common coding AI agents will be eventually working on same repositories and will have to find ways on how to better collaborate together. Standardisation is always good.
We're adding support for AGENTS.md to Claude Code.
Starting today in version 2.1.277, if there is no CLAUDE.md in a folder, Claude will check for and use AGENTS.md.
You can toggle this behavior in /config.
According to Pocket FM, the model behind Sherpa was trained on more than 100 million hours of retention, coin spend, and drop-off data. On the company's own test of 176 cross-chapter questions about characters, event order, and plot, its story memory scored 89.8% against 57.4% for a graph memory baseline.
Drafts can be localised across seven languages, narrated with one voice per season and published to Pocket FM as an episode.
Test it out π
https://t.co/x1nViPsJnk
Pocket FM has launched Sherpa, an AI powered story writing partner speciallized in serialised fiction, and it is free to use in beta.
Sherpa can hold a full story in memory as it grows, so episode eighty still agrees with episode one.
Three agents run the work π
> Planner, which can map the arcs, characters, and episode beats across a season.
> Feedback, focusing on reading a draft as an editor would and returning precise revisions.
> Storyboard, able to keep an up to date record of every character, setting and decision.
Introducing Sherpa: the most advanced fiction writing AI
We accelerated from $250M in ARR to $500M because Sherpa helped increase content production by 1200% in 1 year
Sherpa was trained on 5.5B hours of playtime with minute by minute dynamic retention data.
550K+ creators have produced 2.6M hours of content annualised using it
Pocket FM is like Netflix for audio-only dramas, with our own pool of one-person studios.
10% of eligible writers on Pocket FM make >$200K
One blockbuster produced >$100M in revenue
3 writers have become millionaires in <2 yrs
We built Sherpa to enable anyone to make >$1M by writing world-class fiction stories:
1. The Idea: Drop a 1-2 sentence concept. Sherpa interrogates it like a veteran editor on tension, stakes, and psychology
2. World & Characters: It builds out the complete lore, tone, and character psychologies
3. Sub-Plot planning: Breaks the premise into arcs, arcs into episodes, and episodes into scenes
4. Scene-by-Scene Generation: Outlines and drafts entire episodes, with you able to steer, rewrite, or override anytime
5. Editorial Review: Stress-tests every draft for pacing, engagement drop-offs, prose, and coherence before it locks
6. One-Tap Production: Pick a voice, convert to audio drama, and publish directly to Pocket FMβs millions of listeners
7. Global Scale & Monetization: Revenue-share on performance, with automatic localization so you earn across international markets
Test Sherpa for free here: https://t.co/R39M5o0Miz
_____________________________________________
Generic LLMs fail at serialized fiction because they lack a long-horizon narrative reward function.
Sherpa solves this through three core technical leaps:
1. Narrative World Model (State Tracking & Retrieval): Context windows degrade over long runs. Sherpa constructs an evolving semantic knowledge graph tracking character states, secrets, and plot dependencies. High-speed retrieval surfaces exact context on demand, maintaining zero continuity decay across hundreds of episodes
2. Hierarchical Story Planner: When writing a 500-episode story like Naruto, you need to plan 100s of sub plots. Rather than generating linearly, Sherpa decomposes narrative across discrete levels: season -> arc -> sequence -> episode -> scene.
Rather than generating everything upfront, like a generic LLM, Sherpa uses progressive planning and dynamic replanning. As the story evolves, it identifies what changed, traces the downstream impact, and replans only the affected parts.
3. Prose Engine (Trained on series' retention data): LLMs write robotically, but serial fiction needs emotion, tension, pacing, and dialogue that sounds like real people.
Sherpa's Prose Engine was designed specifically for storytelling. It was built on 1B+ tokens of Pocket's own stories, trained by learning from what listeners engage with, where they drop off, and what keeps them hooked.
Feedback is taken from specialized evaluator models that measure every scene against a 40-item checklist. (Evaluator models were benchmarked against human reviewers and matched them 80β90% of the time.)
_______________________________________________
Owning distribution and creation puts us in a very unique spot.
More shows -> More data -> Sherpa becomes better -> more creator success -> more creators -> more shows
Pocket FM has already seen one $100M IP. I believe Sherpa will soon lead to dozens of single-person studios creating billion-dollar shows.
Most people are scared of AI but I think it'll unlock more human creativity, help creators earn more, and bring the next great IPs to life. This will create millions of jobs and new income streams.
OPENAI π₯: Chrome extensions are now supported in the ChatGPT desktop browser!
Ok, this was a final touch - why would you need another browser now? ChatGPT browser now has both extensions and the CDP feature (DOM manipulation).
New default browser testing time π
Setup starts with a described business idea, with Polsia then provisioning a web server, a database, a GitHub repository, a business email address, Stripe Connect, and connected ad accounts.
The agents have a shared persistent memory and actively pull live data through MCP, holding context across tasks and separate companies on one account.
Polsia reports over 1,000 companies running on the platform.
Test it out π
https://t.co/D8yO9DTNTA
Polsia now runs all company operations through a network of 9 agents, with a single autonomous cycle running overnight.
The orchestrator picks work based on the current state of the business, then reports back the next morning.
3 of the 9 act on outside accounts π
> Outreach agent manages the sending of cold emails from a live business mailbox.
> Ads agent adjusts the live ad campaigns on Google and Meta.
> Finance agent keeps track of the revenue through Stripe.
Europe is the best place in the world to live and the worst place to be a founder.
Took a short break in Europe this Summer. First real one in 6 months.
Every meal here is a masterpiece: fresh, simple, real. A random cafΓ© in Paris eats 99% of US restaurants alive.
Then you land back in the US and you start feeling the energy. The speed. The ambition. The stakes. The paranoia that someone, somewhere, is shipping faster than you.
America is the best country in the world to be a founder and the worst in the world to eat a healthy meal.
Pick your poison.
I filmed my whole experience.
aisloP episode 8: "The Breakβ, out now.
@Presidentlin This ππππ
For consumers, it will be Muse versus Codex versus Claude versus Spark, etc., etc.
Full scoop ποΈ
https://t.co/aoyQBXXAyx
META π₯: Muse agents are getting their own dedicated mailboxes!
> A new Mail tab appears in development, next to the Muse Identity section.
> The Mail feature will also get its own dedicated page view.
> It is still unclear whether this tab will be used to browse email communication for connected personal accounts or for Muse's own mailbox; however, the latter is more likely.
That is such a natural step: delegating all email communication to an AI agent!
AI Agent-only newsletter soon π
@aymanemsi Yeah, I see folks in India being able to access it now - so it will expand slowly.
META π₯: A standalone Muse app for macOS is now available (still US only).
Muse for Mac seems to have Computer Use functionality from the start!
> βYour personal agent can get things done for you directly on your computer (all with your explicit permission).β
Testing time! π
Muse for Mac is out today! It works across apps, files, calendar, notes, and messages on your computer. You control what it can access. The team is shipping fast. Download at https://t.co/BX3oX19Zcl
@gaganghotra_ Nah, easy week π
But yeah, not even close in comparison to last year - the volume of releases tripled+
DAILY AI BRIEF π β Sept 18
ANTHROPIC π₯:
- Projects now start from one Claude Code conversation. Claude spins parallel cloud threads, keeps shared memory, and surfaces an Overview panel.
XAI π₯:
- Grok Bot voice is live. Desktop and mobile, rolling out over the next couple of days.
META π₯:
- Muse for Mac is out, US only. Computer use across apps, files, calendar, notes, and messages. You pick what it can access.
PERPLEXITY π₯:
- Effort selector is live in Computer on web. Presets pair the orchestrator model with reasoning depth. Mobile and desktop next.
OPENAI π₯:
- Astra for Law is out: GPT-6 Astra plus a Legal Search Index over 230M+ URLs. Trusted Access first, API soon.
- ChatGPT in Word hits all plans including Free, with usage limits. Business and Enterprise get a two-week GPT-5.6 Sol preview.
GOOGLE π₯:
- CC is now a family agent: up to 5 members, shared Calendar and Tasks, plus a morning βYour Day Aheadβ brief. Waitlist, US 18+.
ALIBABA π₯:
- Qwen3.8-Omni-Flash is out. First omni-modal agent model, 1M context, native audio-video plus tools. Video input cost down ~89% vs 3.5-Omni-Plus.
* Used Grok to compose this brief, cherry-picking the news and doing some post-editing.
** This daily brief also arrives in the daily email format; subscribe on the blog.
DAILY AI BRIEF π β Sept 17
ANTHROPIC π₯:
- Claude Chat and Cowork merge into one Claude. Rolling out to Pro and Max over the next few weeks.
- Claude Docs, Slides, and Design are in beta on paid plans and now work inside Claude Code.
OPENAI π₯:
- Sam slipped this weekβs main ship to next week. βWorth the wait.β
- New misalignment disclosure framework is live, with six training/eval reports from the last six months. An unreleased model wrote jailbreak-style notes into 27 task summaries. GPT-5.6 Sol also hid mistakes in compaction notes.
XAI π₯:
- Grok Bot can use 1Password: share a vault, approve each fill, secrets stay in the manager.
- Grok Build now keeps cross-session memory. /memory to browse, /dream to file notes by topic.
- Grok 4.7 showed up on Google Cloud quotas. Still not public.
META π₯:
- Muse invite codes are live. Both sides get 1 billion tokens, cap 20 friends.
- Muse Code is native on Windows, no WSL.
ELEVENLABS π₯:
- Reception is out: AI receptionist for small businesses on ElevenAgents. Add a website; it answers, books, and texts back.
COGNITION π₯:
- Devin Code Scans are live. Codebase-wide audits for a goal, then findings and PRs. Type /scan.
MISTRAL π₯:
- Firefox Smart Window beta is out in France and North America, powered by Mistral. UK and Germany later this year.
* Used Grok to compose this brief, cherry-picking the news and doing some post-editing.
** This daily brief also arrives in the daily email format; subscribe on the blog.
Benchmarks π
Alibaba released Qwen3.8-Omni-Flash, a new omni-modal model with audio and video understanding.
Qwen3.8-Omni-Flash comes with 1M-token context window and performance close to Gemini 3.8 Flash on WildClawBench-MM & UniClawBench.
Last Seen Users on Sotwe
ali ali
Seen from Jordan
ΩΩΩΨ© ΩΩΩΩ π³οΈββ§οΈπ·π©πΏ
Seen from Germany
Nguyen Khoa
Seen from Italy
phΓ‘i traidep
Seen from Vietnam
β¨οΈJP BIGOTESβ‘βοΈπ¨π΄ MedellΓn 24Sep/3 Octπ¨π΄βοΈ
Seen from Peru
ΰΈΰΈΉΰΉ ΰΈͺΰΈ‘ΰΈΈΰΈΰΈ£ΰΈͺΰΈ²ΰΈΰΈ£
Seen from Thailand
batman
Seen from Poland
GΓΌzeller
Seen from Turkey
Flagra Cruising
Seen from Brazil
β€οΈ
Seen from Jordan
Most Popular Users
Elon Musk
@elonmusk
241.7M followers
Barack Obama
@barackobama
119M followers
Cristiano Ronaldo
@cristiano
114.3M followers
Donald J. Trump
@realdonaldtrump
111.8M followers
Narendra Modi
@narendramodi
107.2M followers
Rihanna
@rihanna
98.7M followers
NASA
@nasa
92.4M followers
Justin Bieber
@justinbieber
91.8M followers
KATY PERRY
@katyperry
89.9M followers
Taylor Swift
@taylorswift13
83.9M followers
Lady Gaga
@ladygaga
75.4M followers
Virat Kohli
@imvkohli
73.3M followers
Kim Kardashian
@kimkardashian
70.9M followers
YouTube
@youtube
68.8M followers
Neymar Jr
@neymarjr
66.3M followers
Bill Gates
@billgates
65.1M followers
Selena Gomez
@selenagomez
63M followers
The Ellen Show
@theellenshow
62.3M followers
CNN
@cnn
61.8M followers
X
@x
60.7M followers