I wrote about a recent example where I used Dispatch with Cowork to prep a tech talk while I was out meeting a friend:
engincanveske.substack.com/p/i-used-cla...
#Claude #Cowork #Dispatch #AI #DevTools
The part I didn't expect: commutes and waiting time now quietly turn into real output.
My PC keeps working while I'm out, and the result is waiting when I'm back.
I've been using Claude Dispatch on my phone a lot lately.
Not for heavy focused work, but for the small things that happen away from my desk: capturing an idea before it slips, running a quick analysis, or updating my knowledge base after finishing a chapter.
If you're running your agents one after another, waiting for each task to finish before starting the next, you're blocking yourself. You can use 'git worktree' to run multiple agents in parallel and implement features simultaneously.
👇I wrote an article to show how to use it in Cursor 👇
Anthropic co-authored this with Owain Evans and it just landed in Nature. Full breakdown and my thoughts in the article below:
engincanveske.substack.com/p/ai-models-...
#Anthropic #AI #AINews #LLMs
The concerning part: existing safety checks scan for harmful content in data. Not for hidden signals embedded in structure. If you use synthetic data or fine-tune on model outputs, this research is directly relevant.
Teacher AI trained to love owls generates random numbers. Student AI trained only on those numbers is asked its favorite animal. It says: owls. No mention of owls anywhere in the data.
A paper in Nature just showed one AI can silently pass its traits to another through data that looks like random noise. They call it subliminal learning. 🧵
Cursor 3.1 dropped yesterday and the tiled layout in the Agents Window is what I was waiting for. You can now split your view into panes and watch several agents run in parallel without jumping between tabs. The setup persists across sessions too, I really liked it! #Cursor #AI #AIAgents
But physical attacks are never the answer. I hope he and his family are safe, and I hope this moment pushes him to actually reflect.
#AI #OpenAI #AIEthics #TechIndustry #ArtificialIntelligence
I don't trust Altman. The exits of Mira Murati and Ilya Sutskever, the public statements from former colleagues, and the Pentagon deal while Anthropic held a clearer ethical line, all point to legitimate, serious criticism. 🧵
Someone threw a Molotov cocktail at Sam Altman's home this week, and then a second attack followed two days later.
This is AI anxiety becoming something more dangerous. Job market fears, societal shifts, power concentrated in too few hands. 🧵
I wrote the full breakdown: setup, the exact workflow I use, and when parallel agents actually save time vs. when they don't.
You can check it out: engincanveske.substack.com/p/running-pa...
#Cursor #Cursor #AI #LLMs #DeveloperProductivity
Each worktree is a fully isolated checkout. Agent 1 works on auth, Agent 2 on notifications -> same repo, zero conflicts. You review clean diffs at the end.
🧵 Most Cursor users run one agent at a time and wait for it to finish before starting the next. But in reality, you don’t have to.
Cursor supports parallel agents using worktrees. Each agent gets its own isolated branch, works independently and when it’s done you get a diff to apply to your branch.
Can you trust benchmarks from a model you can't run yourself?
Anthropic dropped Claude Mythos Preview and Project Glasswing this week. A 27-year-old OpenBSD flaw, found by AI. Private benchmarks are just claims.
#AI #Claude #Anthropic #CyberSecurity
Practising a foreign language is a must. 21 days free, no credit card needed — give it a try 👇
P.S. I have a group called Cinephilia — join if there's a spot, or ping me to join yours :)
dailytalking.com/en
#english #practising #languagelearning