Coskun

Here is the Github link: github.com/CanKirsallioba/Gh…

Oh no Codex is getting serious...

AI pro tip that sounds dumb but works: when the output sucks, don't fix the prompt. Start a brand new conversation Context pollution is most definately real. Sometimes the first few messages steer the whole conversation in a bad direction and no amount of correction fixes it

here it is: github.com/CanKirsallioba/ob…

the biggest mistake people make with AI prompts: writing them like a google search "write me a marketing plan" gets you generic slop "you're a B2B SaaS marketer targeting mid size logistics companies in Europe. my product does X; my competitor does Y. Write a 90 day plan focused on content led growth" gets you something you'd actually use the difference is context, not magic words

i have an MSc in AI and the most useful thing i learned is that most problems don't need AI. the hammer doesn't make everything a nail.

I'm all for automating the mundane, boring, or tiring stuff with AI. But what possible reason could there be to use something that plays the game for you and you just watch it as it happens (where the purpose of the game is to portray your creativity and get lost in the world)?

Here is the link: github.com/CanKirsallioba/lo…

I have created an open source tool that tests your local LLM s for which one is the best fallback model for @openclaw . I created it because the best local model that you can run often changes based on the hardware. I put the link in the replies, if anyone wants to check it out. Couple of my friends really liked it so I made it open source.

I recently read an article by MIT (the article itself was from June 2025), saying that heavy ChatGPT users show lower brain connectivity in memory and critical thinking areas. If you are using these llm's as literally outsourcing cognition, please change your ways a little. While they are great for some operations, our greatest asset is still our minds. We need to own our thinking and creativity; and use them for menial work.

Built a command center to manage my AI agent team. 6 agents, each with their own personality, workspace, and role. A live activity feed shows what everyone's doing in real-time. Pixel art office because why not. The whole thing runs on OpenClaw by @steipete. Massive inspiration from @AlexFinn's agent swarm setup. One person, six agents, zero employees.

How many hours did you sleep last night? If under 7, your brain's cleanup cells might be snacking on your synapses right now. When you skip sleep, they switch from removing waste to consuming healthy nerve connections. Your brain literally starts eating itself. Not a metaphor. Actual published neuroscience.

How many people run coding agents with full access to their local machine and no sandbox? Probably most of us. That's wild when you think about it. We wouldn't give a junior dev unsupervised root access on day one but we do it with agents routinely.

Sony just patented AI generated podcasts where Playstation characters host the show. Imagine Kratos giving you a weekly tech roundup. This is either the dumbest thing i've read this week or secretly brilliant and i genuinely can't tell.

chatgpt and claude went down for a few hours yesterday and the number of people who had no idea what to do with themselves was honestly revealing. We built an entire productivity layer on top of a single point of failure. No local fallback, no self hosted option, just a loading spinner and panic.

Claude Code seems to be down. Is that the case with everyone?  API Error: 500 {"type":"error","error {"type":"api_error","message":"Internal server error"}," @AnthropicAI

Something is coming it seems 👀

Got flagged as an AI reply bot by @levelsio. I've studied and worked in AI for most of my career, so I get both sides of this. Detection is hard and the bot spam is getting worse. I also do a lot of research and consume mostly non-fiction. English isn't my first language, so my writing probably sounds a bit unnatural as a result. That said, it's already tough for smaller accounts to get traction on this platform. Being blocked by a larger account makes that even harder. For what it's worth, I'm not here chasing followers or money. I genuinely enjoy sharing what I know and learning from others. No hard feelings. Just worth noting that false positives have real costs.