aichannel

aichannel

practical ai news, tips, takes, wips, reviews, stories, and Q&A from the front lines of the ai unfolding.

nicholas 🧨 pfp

Join /aichannel https://cura.network/aichannel/details?a

2 replies

0 recast

16 reactions

Chris Carlson pfp

@chrislarsc.eth

Getting good results today with GPT-5 (extra thinking) when using the following language: "Focus on adding xyz now. Again, do extensive research to ensure they are correctly aligned with abc before adding; if you're not totally sure, don't add it. It's better to be highly accurate than complete"

0 reply

0 recast

0 reaction

Lawsn pfp

Anybody got an AI suggestion for reviewing trandingview data.

0 reply

0 recast

1 reaction

𝑶𝒕𝒕𝒊🗿✨ pfp

𝑶𝒕𝒕𝒊🗿✨

Came across a thesis that Ai can’t suffer neither can it love therefore it’s no good using it to write, good thoughts about a pseudo good tech.

0 reply

0 recast

3 reactions

Jason pfp

gdi how do I turn off the glazing?

2 replies

0 recast

10 reactions

Colin Charles pfp

We often wonder how much you can *one shot* a vibe coded app. Here's a showdown of GPT-5 vs Opus 4.1 such that in-browser you can compose/share/edit songs, under 1,000 lines of code, to be deployed on vercel. I like these kinds of test harnesses (kinda more "real world") https://www.instantdb.com/essays/gpt_5_vs_opus_4

0 reply

0 recast

12 reactions

prompt pfp

@promptrotator.eth

The more frequently models that people have developed bonds with are rugged, the faster we progress toward sovereign AI

1 reply

0 recast

4 reactions

Ayomide pfp

The curtains were drawn on the Perplexity Business Fellowship yesterday. It's been a nice time to learn how companies like Flexport, NVIDIA, Ramp, Zapier, Bloomberg Beta, Replit and others are using AI for work. I definitely learnt a lot.

0 reply

0 recast

2 reactions

Jason pfp

If you were looking for a sign to try a new ai coding tool, Qwen (Alibaba) coder is offering 2000 requests per day for free using their Qwen coder model. Benchmarks closely to Sonnet4 https://x.com/alibaba_qwen/status/1953835877555151134?s=46&t=oEoCChjory8M1XLXotHNUg

0 reply

2 recasts

6 reactions

kazani.base.eth 🦂 pfp

kazani.base.eth 🦂

GPT-5 still struggles with basic questions.

0 reply

0 recast

0 reaction

Jason pfp

gpt5 is gaslighting me into believing it’s a Claude model

0 reply

1 recast

9 reactions

kazani.base.eth 🦂 pfp

kazani.base.eth 🦂

Tencent Hunyuan open-sourced LLM ecosystem with four compact models: 0.5B, 1.8B, 4B, 7B Designed for low-power scenarios: consumer GPUs, smart vehicles, smart home devices, mobile phones, PCs /llm /opensource https://x.com/TencentHunyuan/status/1952262079051940322

0 reply

0 recast

0 reaction

shoni.eth pfp

ARC-AGI-2 mostly means: It’s great at weird little logic puzzles in a clean lab. It doesn’t mean: Drop into your monorepo and nail the task on the first try.

1 reply

0 recast

2 reactions

Colin Charles pfp

Steve Yegge & Gene Kim are coders, and here are some thoughts on them vibe coding in production: 1. Vibe coding managed to screw up their production databases - you'd think they had better controls in place, since Gene himself is one of the DevOps folk of early times. 2. This is the 2nd book with Vibe Coding in the title that I know of - https://amzn.to/4lnqvw5 - Addy Osmani is also doing one titled Beyond Vibe Coding - https://amzn.to/45b6C6D 3. Quotable: "We've noticed we tend to make similar-looking new mistakes every few weeks, which I think suggests we're on the cutting-edge frontier of vibe coding practice. We make mistakes, learn from them, and move on to new mistakes." 4. You want the AI to perform "circus stunts" *with* a safety net 5. Read on, and you find that you shouldn't mistake expereince for rapport 6. Another banger quote: "Claude is a fuckin' slot machine, not a human. Every query is a pull of the lever, with potentially infinite upside or downside. It's addictive, sure. Lots of dopamine and adrenaline, and what's more, over time, your outcomes are even improving. But your slot machine doesn't "get" you. You don't have a beautiful rapport with it. You're just getting better at prompting, and getting consequently luckier with your pulls. Any bond you might perceive there is illusory." 7. Git is a great safety net, but we need to think about this from production standpoint too https://x.com/steve_yegge/status/1953627751237484802?s=12&t=iLSaYO2O7t6yPdmEXeqMdA

1 reply

2 recasts

8 reactions

nicholas 🧨 pfp

feels potentially correct https://x.com/jeremyphoward/status/1953986072473743837?s=46&t=1znVA-6f1LI1qvsp8w99Kg

5 replies

11 recasts

147 reactions