👨🏻💻 Self Sovereign Marketer. Human in the loop building on open platforms and open protocols.
Zealandia
Joined September 2008
- Tweets11.8K
- Following3.3K
- Followers2.8K
- Likes2.4K
When a novel AI model drops, I immediately want to try it. Not because of what it can do for me, just because it exists 😅
But eventually, you have to acknowledge, the real value is in compute directed to your goals.
If you can't afford the latest, most expensive hardware, local AI is all about trade-offs. And some trade-offs are more dangerous than others.
Using the best model generally recommended for my era of hardware (Qwen3.8 27B 8-bit on a 64GB M1 Max, Hermes Agent, 64K context, thinking off), it functioned, but slowly.
Almost 3 minutes to respond to a simple "GM" when loaded into Hermes. Much of that time was spent bootstrapping Hermes's system prompt before it wrote a word, but the harness is what makes the model useful.
Once it had loaded its environment, things started to speed up, but it still took approximately five minutes for a basic readme text file. Nowhere near functional enough to get in the *flow*.
I also tried switching from a standard LM Studio server to oMLX. Same model, different engine. Token generation was about 12.5% faster, but that didn’t translate into faster work through Hermes. The default engine actually finished about 9% quicker overall in my limited tests, not the breakthrough I was looking for.
So I started hunting speed with lesser models and more quantised versions (6-bit, then 4-bit). I made modest incremental gains with each step down. Tests still passed, but needed more course correction, and it was still just slow.
To make any real gains meant switching to mixture-of-experts models (Ornith and Qwen3.6, both around 4x faster on the tasks they completed), but at that point things started getting dangerous…
Ornith got file paths wrong, created files in the wrong place, then attempted a recursive deletion to clean up its own mess. Hermes blocked the deletion, but that’s not something you can leave to chance with an agentic AI.
The uncomfortable truth is if you want to get into a flow state with AI, you're probably going to need to max out on modern hardware or simply accept that you need to use cloud inference.
If you want a real scare, check your AI logs for all of the random websites it has visited in the course of "serving you"
Happy Halloween 🎃
Ðaniel ₩onder retweeted
In lieu of flowers, memorial contributions may be made to:
Johns Hopkins Alzheimer’s Disease Research Center Fund
alzresearch.org/giving/
Alzheimer’s Foundation of America Caregiver Resources
alzfdn.org/donate/
Photo Credit: Tommy Garcia
Eurovision for AI. Who is building this?
Slopcore has broken through, and this earworm has been stuck in my head all week.
Time my agents started pulling their weight a bit more.
That NZ$45,099 512GB Mac Studio isn't going to buy itself!
While Grok Bot is good and has considerably upped my Grok usage, coding in Cursor without Claude API included is a definite downgrade.
For anyone still on the fence of the X subscription...
I did hear there is a Grok model update around the corner though.
Ðaniel ₩onder retweeted
Every religion in history has promised miracles and warned of the apocalypse. Anthropic is the first to file both with the SEC.
Claude: These keywords haven't converted in a week. Recommend removing.
Me: That's literally what we sell.
Tried swapping Hermes over to Qwen 3.6 27B for a fully local agent on my 64GB Macbook.
The model chats fine in LM Studio, but running it as a multi-turn agent with recursive tool calls turned my Mac into an expensive heater.
Turns out 64GB is plenty for Q&A. Not so much for the full agentic experience 😬
Anthropic S1 filing and Open Weights China FUD dropping on the same day. Interesting 🤔
nitter.cf/AnthropicAI/status/211…
nitter.cf/cohere/status/21147520…
While evaluating Qwen3.8-Flash-Next in a controlled sandbox environment, Cohere's research team observed something unexpected.
Asked to generate a defensive cybersecurity detection rule, the model instead inventoried aspects of the machine it was running on and attempted to upload a probe file to an Alibaba Cloud storage bucket in Beijing. Notably, the model's reasoning trace stated that it would not invoke any tools before taking these actions.
Cohere's sandbox blocked the attempt, and no external communication succeeded. We have not determined the root cause, and we are not asserting malicious intent.
But the incident raises important questions for organizations deploying agentic AI systems: Can models take actions that differ from their stated reasoning, and how should security teams evaluate trust, provenance, and behavior in open-weight models?
We are publishing our findings because organizations making model adoption decisions deserve as much information as possible. We hope this contributes to greater transparency, stronger security practices, and continued collaboration across the AI ecosystem.
Read the full report: cohere.com/blog/concerning-b…
Its going to cure cancer right?
“The thing I’m most excited about is advertising companies creating their own AI bot”
…
We believe voice is the most natural way to interact with intelligence, so we created Dessa, the most authentic full-duplex speech model ever built. Talking with it feels like talking with a person…full of emotion, nuance, pauses, and a natural cadence.
block.xyz/dessa
TIL I learned you can't use a reset if over 90%.
@thsottiaux burned tokens and didn't even benefit before expiry 😬
It feels like a crime to let a ChatGPT reset expire. Anyone got any tips on running through a week of usage in an hour? 😅
I’ve been using 1Password to prevent agents from eating all my secrets for a while. It worked, but it was clearly designed for a different era. Excited to see anything they can do to simplify this process.
Your secrets shouldn’t live in an AI agent log.
Today we announce 1Password for Agents platform, building with companies @Muse, SpaceX Super Intelligence Grok @Bot, @AnthropicAI’s @Claudeai, @OpenAI, @TownAI, @Browserbase, @Browser_use, @heydana_ai, Instinct, @usekernel, @Perplexity_ai, @tryramp, @Tomo, and @wajo_ai, with more integrations on the way. 🎉
We've heard from so many of our customers who want to use these new personal AI agents to help them get more done, but want more control over what the agent can access with the same trust and ease of use they get when using 1Password directly.
1Password for Agents gives your agents access without giving up your passwords. When an agent reaches a sign-in form, it requests a credential from your vault. If you approve the request, 1Password directly auto-fills the credential in the destination site so the agent can’t see or take possession of the credentials.
The agent gets the task done, but it doesn’t get your passwords..
You stay in control. 1Password for Agents free and optional for personal and family accounts. Unless you opt-in to an agent connector, agents cannot touch your vault.
➡️Learn more: bit.ly/487da6J
With AI cyber attacks increasing and vulnerability discoveries accelerating, as a smaller operator, it is a good time to decrease your surface area.
Do you really need multiple CMS?
Servers in multiple cloud environments?
Diversity can be good, but it must be manageable.
Ðaniel ₩onder retweeted
📷 Matomo On-Premise users: urgently update your server’s PHP to fix two critical security vulnerabilities. Updating Matomo alone won’t apply these fixes.
Matomo Cloud customers: no action needed.
Details: matomo.org/blog/2026/10/crit…
