Armarium 1.1 is out π
Now free to download. Try everything for 7 days, then unlock it once and it's yours. Works with Family Sharing. No subscription.
Bought 1.0? Nothing changes for you, everything stays unlocked.
New in 1.1:
β’ Keep on Device: an iCloud book stays in the app and plays offline
β’ Series: when a book ends, Armarium offers the next volume
β’ Bookmarks can save a range, like your favourite 30 seconds
β’ Steady volume: loud recordings are turned down so books sound alike
β’ Covers from a cover image next to your files, straight from iCloud
β’ Apple Watch: send queue, and speed up to 2.5Γ
Plus a bunch of fixes for folder books and chapter files.
Your app. Your books. Your rules.
https://t.co/fVstmAIYXT
I moved my gaming PC from Windows to Linux. An AI agent did most of the work. It still wasn't easy π§
Day one my Xbox controller kept dropping out. Drivers? No. The Bluetooth dongle loads its firmware from the OS, and Windows ships a newer one. The fix was free: one file from Microsoft's own update catalog, copied into Linux.
Then it hung again. I'm still fighting it.
Meanwhile: 916 GB of games backed up before the wipe, saves moved by hand where Steam Cloud doesn't reach, and now the PC boots straight into Steam like a console. No password, no desktop.
The agent researched, read logs and wrote the commands. I typed every sudo myself.
Best part: one tap on my Apple Watch wakes the PC and the TV, dims the living room to amber and starts The Witcher 3 RemasterπΊ
Thread with the fights coming π
I built the audiobook player I always wanted π
Armarium plays your own mp3, m4b and m4a files on iPhone and Apple Watch. Leave the phone at home, put on AirPods, listen from your wrist.
No store, no account, no cloud. No subscription either.
Free to download, 7 days to try everything, then you pay once and it's yours.
Your app. Your books. Your rules.β¨https://t.co/fVstmAIr8l
And if your e-books have messy titles and missing covers, I made Titulus too π·οΈβ¨https://t.co/eS0Uo7OQqT
π€ AI in brief, 12 Nov 2026
π£ SpaceXSI (formerly xAI) posted Grok Bot 1.0 on 11 Nov: Bots can be shared and are "more collaborative and capable than ever." The newsroom page, dated 2 Nov, says you can share Bots and bring them into Slack, Microsoft Teams, and video calls. https://t.co/caJep6zost https://t.co/OUCCr5Hw2s
π£ Claroc tried 43,551 Grok app addresses. Of 37,571 still online, 35,578 (94.7%) opened with no sign-in. 24 were built by SpaceXSI and X staff, including a confidential onboarding plan and a sales map of 594 accounts. In Claroc's account, Publish proposed "Anyone with the link." SpaceXSI says its employees' apps start out private. Claroc says SpaceXSI took those 24 down after the report, but not the apps other users left open. https://t.co/xdWHCm7yNm
π΄ Elon Musk, quoting @bot on 11 Nov: "Hire Grok @Bot as a colleague and give it your hardest tasks!" https://t.co/LIJW9uPamR
π‘ The Wall Street Journal reported on 11 Nov that Nvidia's board is preparing to oust CEO Jensen Huang after two extramarital affairs, one with a suspected Chinese spy and one with an employee. CNBC said it confirmed late Wednesday that the board was meeting on whether to remove him, and that shares fell more than 9% after hours. CNBC, citing the New York Times, named healthcare VP Kimberly Powell. https://t.co/s7dbAsoGCO https://t.co/onVsDURjBK
π‘ Google launched the Gemini Decisions API on 11 Nov. One request scores up to 1,000 yes/no, choice, or rubric questions. Price: $0.10 per million input tokens, $0 for output. Google says latency is 150 to 250 ms and it's nearly 5x cheaper than Gemini 3.5 Flash-Lite on its benchmark suite. https://t.co/Ru721Wv3Uu
π‘ DeepMind released AlphaGenome 2 on 11 Nov. It predicts how DNA variants affect specific human cell types. On 28 immune cell types it identified nearly twice as many causal variants (30% vs. 16%) as the original AlphaGenome fine-tuned on single-cell data. Not approved for clinical use. https://t.co/ZBgT96kxWo
#AI #Nvidia #Grok
π€ AI in brief, 11 Nov 2026
π Anthropic launched the roadshow for its proposed IPO. The notice offers 825 million Class A shares: 801,351,354 from the company and 23,648,646 from existing holders, expected at $70 to $78. Anthropic expects to give underwriters a 30-day option on up to 123.75 million more. Proposed Nasdaq symbol: ANT. The registration is not yet effective, so nothing is on sale yet. https://t.co/g7nhNl2Esw
π£ SpaceXSI (formerly xAI) hasn't posted about it, but Elon Musk says the team dumped JAX and SGLang and rewrote the RL and inference stack in C/C++. Grok 4.8 pre-training is done, the model is in RL, evals are starting to climb, and a soft rollout to Bot users is planned for later this week. Musk wrote that 4.8 "will bear no resemblance to 4.7" and "is not yet incredible." https://t.co/Ors6XOOnus
π΄ Asked how 4.8 will compare with Claude Opus 5.5, Musk wrote: "Not as good. Opus 5.5 is a masterpiece." He added that Grok 5 should exceed Opus 5.5, "and maybe that happens in 2 or 3 months." https://t.co/kzbx4NNeHU
π’ Bloomberg reports OpenAI's revenue hit $10 billion in the quarter ended September, its first at that level, nearly 50% above the prior quarter and almost threefold a year earlier. Bloomberg's sources are people familiar with the matter; the figures were shared with investors on a Tuesday call and are not public. https://t.co/mJd6qpHHTB
π’ Businesses in ChatGPT is in beta. A company can claim its profile, update details, add product feeds, and see aggregated mentions and clicks. Eligible US ecommerce and online-service businesses can add a Business Agent. Its chats are clearly labeled and kept apart from ChatGPT's own answers, and OpenAI says it doesn't share them with the business. https://t.co/MttytC7lH8
π Claude Whiteboards and Claude Sheets are in beta on paid plans. Claude Dashboards is now generally available as Claude Data, with certified queries and live sharing still in beta. https://t.co/xdurX21zTJ
#AI #Anthropic #Grok
@Dimdv99@tetsuoai Not as good. Opus 5.5 is a masterpiece.
But Grok 5 should exceed Opus 5.5 and maybe that happens in 2 or 3 months, with continued acceleration thereafter.
A week ago the answer to Grok Bot's usage limits was "upgrade".
Nov 1, Elon Musk, as users burned through their weekly allowance: "The cost of serving Grok Bot was unsustainably high relative to revenue." To someone locked out until Wednesday: "Upgrade to higher usage plan cost increase is too high?"
Today it's a different tune π A round table with the Grok Bot engineering team, fixes going in overnight, and "you should notice a dramatic improvement by tomorrow". If not, reply and they'll address it right away.
Good that the feedback finally landed. But read what the post doesn't say: nothing about weekly limits, nothing about prices. "Improvement" can mean a smarter, more efficient agent, not more usage for your money.
So Tuesday is the test. I'll believe it when my usage meter shows it.
For me the limit already settled it: that cap is exactly why I moved my daily automations off Grok Bot to Hermes Agent this weekend, running on my own home server. Curious whether Tuesday brings people back.
https://t.co/ghyfqqQrfz
https://t.co/yIdCQxH5gT
#Grok #AI
Just did a round table review with the Grok @Bot engineering team.
Fixes are being implemented overnight and you should notice a dramatic improvement by tomorrow. If not, reply to this post and we will address any remaining issues right away.
π€ AI in brief, 9 Nov 2026
π‘ π‘ Nvidia's board opened an independent investigation on Sunday after reports that Jensen Huang was the target of a foreign intelligence effort. The board said it has no reason to believe national security or Nvidia secrets were compromised, that Huang fully supports the review, and that it still has confidence in him. Out-of-hours Nvidia was about $180.30, down 3.4% from Friday's $186.64 close. South Korea said contracted GPU deliveries should not be disrupted.
https://t.co/K6N6uHRUk3
π‘ NVIDIA, Rebellions and SK Telecom, backed by Korea's science ministry, will connect Rebellions' R200 NPUs to NVLink Fusion for rack-scale inference. SK Telecom plans to deploy them. The systems are aimed at 2028, not this year.
https://t.co/3jhd1XuZ2S
π£ SpaceXSI (formerly xAI): Musk said he sat with the Grok Bot engineering team. Fixes go in overnight, and he wants a dramatic improvement by Tuesday. If it does not show, he told people to reply on the post. No new price and no new limit.
https://t.co/ghyfqqQrfz
π΄ Musk replied "Yup" to a paraphrase that 100 GW/year of orbital AI compute at $10 a watt would add $1 trillion a year to SpaceX. That figure is the post he affirmed, not a filing. On Huang he replied "I agree" that Jensen did something dumb and bad for America, and is still hard to replace.
π‘ Trump told CBS's 60 Minutes he does not mind AI guardrails, as long as they do not put the US in "a deep second position" behind China. Reuters also reported data-center buyers hunting used gas turbines because new ones are being quoted for 2030-31. A new US turbine now averages six years. Worldwide orders top 100 GW against 60-70 GW of factory capacity.
#AI #Nvidia #Grok
Just did a round table review with the Grok @Bot engineering team.
Fixes are being implemented overnight and you should notice a dramatic improvement by tomorrow. If not, reply to this post and we will address any remaining issues right away.
@plainandsimpl_ Armarium reads embedded m4b chapters first, then a .cue file with the same name next to the file. If neither exists, the book plays as one long track. You can still seek, skip by seconds and drop named bookmarks anywhere in it. No auto-splitting, and your file is never modified.
I built the audiobook player I always wanted π
Armarium plays your own mp3, m4b and m4a files on iPhone and Apple Watch. No store, no account, no cloud.
Chapters, bookmarks, sleep timer, widgets, Siri. One-time purchase.
Your app. Your books. Your rules.
https://t.co/fVstmAIYXT
@unflailinglove@NousResearch Interesting! Which model exactly were you running in Hermes? In my experience the model choice changes the tone a lot, some of them feel way more robotic than others.
For the last few months my AI assistant was Grok Bot, running 24/7 on my home server. It did its job: routines, research, reminders.
This week I moved everything to Hermes by @NousResearch. Three days in, I already know I'm not going back.
The move took one evening. Hermes imported my Grok Bot export with memory, 26 skills and routines, so I didn't start from zero. From the first chat it knew who I am, what hardware I have and which rules to follow.
The biggest change: instead of one provider, I now run several bots on different providers. Each one uses a different model and does a different job. So when something like a @grok limit hits, the others keep going.
#AIagents #selfhosted
Not everyone knows this, but Hermes by @NousResearch has /compress.
Long chat getting heavy? Type /compress. It summarizes the middle of the conversation, keeps the start and the latest messages, and continues in a new session.
Saves tokens, keeps the context.
#AIagents#AI
@24thIDRefugee No local models here, so no beefy GPU needed. Hermes runs on my home server, but the models themselves run in the cloud through providers like Gemini and Grok. The agent part is light, almost any machine or a cheap VPS can handle it.
@unflailinglove@NousResearch Different experiences, and that's fine. Everyone should pick what works for them. For me Hermes has been great. The tone depends a lot on the model and the SOUL.md personality file, so mine sounds nothing like a robot. But if Grok Bot clicks for you, go for it.
100% agree. Self-hosting isn't automatically safer, it just makes security your job. My basics: agent runs as a normal user, not root. Risky commands need my approval. Nothing exposed publicly, only Tailscale. And I treat anything it fetches as untrusted, because prompt injection is real.
@MathiasMei5r@NousResearch Same here, same reason. Those limits were driving me crazy. Moved last week and haven't missed Grok Bot once. Nice to see it worked out for you too π
π€ AI in brief β November 6, 2026
OpenAI unveils a cybersecurity model, Claude moves into personal finance, and AI spending worries markets.
π’ OpenAI is releasing GPT-6 Cyber to selected customers, rating its cybersecurity capabilities as βcritical.β It helped find 81 new serious Firefox vulnerabilities and write 948 Linux exploits.
π’ OpenAI pledges to cover the full electricity costs of its future data centers, publish construction plans, and avoid NDAs with public officials.
π Anthropic launches a US Personal Finance beta: Claude gets read-only access to bank and card accounts through Plaid.
π The Register: a researcher claims a Claude Dispatch flaw lets a browser extension execute code on a computer, and says Anthropic has not fixed it.
π£ Grok Bot can now make phone calls. Starting November 7, Grok API text will also carry an invisible watermark.
π‘ Nasdaq has fallen more than 10% from its peak, entering correction territory amid concerns over AI spending and data-center regulation.
Sources:
https://t.co/WKTjG0Wjy6
https://t.co/yuQEBBnMjz
#AI #OpenAI #Anthropic
@freshfey@NousResearch Neither actually. Hermex by @uzairansar is open source, so I forked it and let my agent on Opus do a full code review. Cut the stuff I don't use, added what I was missing. A few Cursor agents, two review rounds, and half an hour later I was testing my own version on my iPhone.
@uvince@NousResearch Fair point, βGrok botβ was just shorthand. The model itself obviously runs on xAIβs servers. What runs 24/7 on my home server is Hermes Agent (by @NousResearch Research), which talks to Grok through the API. So itβs my own bot, on my own hardware, with Grok as the brain.
@ifexy911@NousResearch That's the nice part: you choose the speed. One of my bots runs on Gemini Flash and replies almost instantly, so it handles the quick stuff. The heavier tasks go to bots on stronger, slower models.