ugur turan
271 posts


80 hours of AI pair programming. Here's what context-mode saved me.
→ $487.20 in API costs. Opus pricing. Real money, not estimates.
→ 22.3 hours of re-explaining context after compaction. Time I got back to ship.
→ 268 sessions resumed from memory. The agent never asked "what were we doing?"
→ 47 preferences auto-learned. "use TS strict" once, remembered forever.
→ 14,847 events indexed. Searchable across every session, every project.
Without context-mode |████████████████████████████████████████| 6.2 MB
With context-mode |█░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░| 124 KB
98% of raw data never entered my conversation.
That's a 50× longer session. Same context window.
Multiply across a 13-engineer team:
$487 × 13 = $6,331/month saved
22 hours × 13 = 286 hours/month recovered
Open source. Local-first. No telemetry. No account. No SaaS lock-in.
github.com/mksglu/context…

English

@Parul_Gautam7 Im still waiting to hear from them to show me my call logs to Kimi 2.6. It’s been days, no reply.
English
ugur turan retweetledi
ugur turan retweetledi

Things are moving fastt!!
Just 12 hours after @AnthropicAI's hackathon, I found myself on a WhatsApp call with @mnedoszytko, who placed third in the previous hackathon with PostVisit AI — which was later acquired by Abridge.
What a humble, warm human. Two coding doctors nerding out for 20 minutes — he generously shared his experience, what comes next, and where Medkit can go from here.
Fun fact: my original idea was a doctor-patient voice assistant. Then I saw Michał's work and thought "he's already nailed that" — so I pivoted to medical education. That's how Medkit was born :)
Truly excited to see what comes next!
Big shout out to @claudeai, @claudedevs, @cerebral_valley & @nosajab for the incredible community 🙌
#BuiltWithClaude #ClaudeCode


English

@MillieMarconnni When you learn to use an app and the ui changes, you say wait a minute, that’s not right and explore. I wonder what this does.
English

A developer just open-sourced an AI agent that learns new computer skills by doing the task once, then remembers the workflow forever.
It’s called GenericAgent.
And the idea is insane:
Don’t preload skills.
Let the agent grow them.
Most agents ship with a giant pile of built-in features you will never use.
GenericAgent starts with a tiny seed.
~3K lines of core code.
9 basic actions.
~100-line agent loop.
Then it takes control of your actual computer.
Browser.
Terminal.
Files.
Keyboard.
Mouse.
Screen vision.
Mobile apps through ADB.
Here’s the wild part:
When it completes a task, it turns the execution path into a reusable skill.
First time you ask it to send a file through Gmail:
It figures out OAuth.
Writes the script.
Tests the flow.
Saves the skill.
Next time?
One-line invoke.
This is not a chatbot.
This is an AI worker that builds its own operating manual while it works.
MIT License.
100% Open Source.
github.com/lsdefine/Gener…

English

1 laptop.
1 Claude account.
1 winning prompt stack.
= $10k/month landing freelance clients on autopilot.
I compiled my entire client-getting workflow into a 38-page guide.
Was going to charge $197.
For 24 hours → completely FREE.
Like + comment 'Need' and I'll DM you my step-by-step guide for FREE.
You must be following me to receive the DM.

English

I spent 6 months learning Claude Code the hard way.
Every mistake.
Every workflow that did not work.
Every prompt I had to rewrite 10 times before it clicked.
I turned all of it into a free masterclass:
THE CLAUDE CODE MASTERCLASS.
Zero to shipping your first project alone.
No CS degree. No team. No guesswork.
Inside you get:
- The reframe that changes how you use Claude Code forever
- The CLAUDE .md template that makes every session 10x more powerful
- The 4-layer prompt architecture that scales to any project
- MCP server setup nobody is teaching yet
- The 6-phase sequence to go from idea to deployed in a weekend
- 10 ready-to-run workflows you can copy right now
The people who read this tonight will be building things next week that most developers still cannot do with a team.
I should be charging $199 for this.
It is free.
Comment CLAUDE and I will send it to you directly.
RT if you know someone who needs to see this.

English

@Teknium @NousResearch Sent you a dm. Claude warned me to alert nousresearch about the web search queries but I couldn’t reach anyone, security mail returned.
English

@viperciyes @NousResearch It clearly says it did interrupt
It did a bunch of parallel tool calls that finished and had to be reported. I made it so that it will just throw those away and never show them again tho so you can feel better ^_^
English

Hermes Agent tip of the day:
There are 4 ways to deal with the model while its running,
- Message it, by default, it will interrupt the agent loop, stopping it and making it respond to your new message
- /queue will queue up a message that will fire after the agent loop completes
- /bg or /btw will run a parallel prompt that is async
And
- /steer will inject a guidance message into the next tool calls result sent to the model during an agent loop, to try to guide the rest of its trajectory
English

@Teknium @tymrtn @NousResearch Nope, it decided to retry on its own. Was using Kimi 2.6. After /stop it did actually stop but the searches were suspicious, nothing to do with my current session. Was it Hermes or the model, no idea.

English

@viperciyes @tymrtn @NousResearch fwiw it looks like it did interrupt and then decided you still wanted it anyways - what model was this?
English

@tymrtn @Teknium @NousResearch Did that in the end. But wanted to show that it says interrupting but it couldn’t, it knows it’s not doing the right thing, tries to fix without confirmation, then does the wrong thing again. Couldn’t break the loop.
English

@viperciyes @smoermeow @Teknium @NousResearch 🤣 mine does this too. It doesn't stop unless I actually didn't want it to stop then it stops.
English

@smoermeow @Teknium @NousResearch Yep, how it stopped in the end. Then I said hi and it went back to web searches
English

@Kimi_Moonshot @NousResearch More context: asked to run simple terminal commands but the agent ran multiple web searches. kindly asked it to stop to find it looping, so did a hard /stop. when I asked why, it apologized and retried, again websearching instead. @NousResearch security mail returned to me.
English

is it @Kimi_Moonshot Kimi 2.6, @NousResearch Hermes, or any config I can't fix but I'm getting agent smith vibes...

English
ugur turan retweetledi
ugur turan retweetledi

Şubat ayında Vodafone sözleşmem doldu. Yüksek fiyat verdikleri için hattımı Turkcell’e taşıdım.
Vodafone tarafında 25 gb internet paketim vardı. Turkcell tarifelerinden de yine 25 gb seçtim. 2 sene boyunca Vodafone kullanımımda paketi %80 bitirmişliğim dahi yokken, Turkcell geçişiyle birlikte bitirmediğim ay olmadı. Bu ay tarife yenilemeye 18 gün varken %48 gitti bile.
Kullandığım cihaz aynı, uygulamalar aynı, kullanım alışkanlıklarım aynı. Ne hikmetse Turkcell paketi su gibi gidiyor. Daha da ilginç bir bilgi vereyim. 10 gün yurtdışı programımda, gittiğim yerden 10 gb internetli e-sim aldım. Bitiremeden döndüm. Burada ne yapıyorsam, nasıl kullanıyorsam orada da aynı şekilde kullandım.
Bize bu durumu açıklar mısınız? @Turkcell

Türkçe

@ASalvadorini @AdamMusaAly Maybe you should improve minimal with everything getx does, so people would migrate 🤣
English

@AdamMusaAly You can't migrate the whole app, as GetX does all of it. But I could write about migrating the state management part. I'll think about it 🤔 thanks for the suggestion
English
ugur turan retweetledi

@kentcdodds I feel the frustration but a skilled (or better paid, not sure) flutter developer wouldn’t release it that way. There are ways around all these things, not flutter’s fault if not implemented.
English













