I had built a personal tool to help me have extremely long conversations with LLMs in my research and analytical projects. These threads got long. Very long. About half a million tokens with Claude and GPT /Extreme%20Thread%20Length/ChatGPT_Thread_450k_tokens-Redacted.md) and over a million with Grok /Extreme%20Thread%20Length/Grok%20Thread%201M%20tokens-%20Redacted). All coherent, clean, and well-reasoned threads with no meaningful drift, hallucination, sycophancy, or other issues that make l...
A complete AI agency at your fingertips - From frontend wizards to Reddit community ninjas, from whimsy injectors to reality checkers. Each agent is a specialized expert with personality, processes, and proven deliverables.
Comments URL: https://news.ycombinator.com/item?id=48756547
Points: 1
# Comments: 0
Synaplan is Apache2, comes with Open Source helm charts for K8s and has all major AI APIS integrated, including Ollama for local fun. Pretty powerful for hosters, who want to run their branded version of it (explicitly wanted).
Comments URL: https://news.ycombinator.com/item?id=48757444
Points: 1
# Comments: 0
Every AI tool now wants to help you shop (ChatGPT, Perplexity, Gemini, Alexa for Shopping). I keep trying them and still default to plain Amazon. Curious where this community lands: has any of it genuinely changed how you buy things, or is it still a novelty you don't reach for? What works, and what makes you not trust it?
Last week I shared the early architecture for IONS, a protocol built around the idea that intelligence should emerge from traversing reusable Cognitive Building Blocks (CBBs) rather than being compressed into increasingly larger model weights. The feedback was excellent. The biggest criticism wasn’t the premise. It was scalability. That criticism turned out to be justified. Today the Genesis network contains roughly 9,000 Cognitive Building Blocks, more than 50,000 relationships, and two f...
A week ago I posted that the Legion LegalTech case in D.C. was testing whether Commerce can treat access to a hosted frontier AI model as an export-control issue. https://www.reddit.com/r/artificial/comments/1uexdqk/followup_hosted_ai_export_controls_are_now_being/ Which was a followup to an initial post about the export control restrictions imposed on Anthropic: https://www.reddit.com/r/artificial/comments/1u4yjdi/does_commerce_have_the_authority_to_apply_export/ There’s now a new wri...
Hi Guys, Don't get me wrong, AI is phenomenal, but right now it still feels like an optional novelty or a niche tool for most everyday folks. To me, it hasn't hit that "can't live without it" status that the internet or smartphones have. Looking only at consumer products (not B2B or corporate software), how do you picture AI being integrated into our lives in the near future so that it becomes a true, indispensable utility? What’s the "killer feature" or shift that takes ...
ServiceNow's customer chief called tokenmaxxing an AI hype cycle in the Observer last week. Not some startup founder flexing contrarian creds — the guy who runs their customer org telling everyone the wrong meter is running the show. Salesforce announced $2 per resolved issue last Wednesday. Same direction, different side of the negotiating table. Stop measuring how much work you did. Start measuring what actually got done. I've sat in enough procurement calls to know why nobody moves on t...
I've been thinking about this a lot lately. There's a real tension between using AI as a learning tool versus using it as a shortcut that bypasses learning entirely. When I use something like ChatGPT or Claude to understand a new concept, sometimes I come away genuinely understanding it better than I would have from a textbook. Other times I just copy the output and move on, having learned nothing. The question is whether that's a problem with AI itself, or just human nature meeting a new to...