Genco Divrikli

Strategic AI Governance Leader | Bridging Legacy IT with GenAI & Agentic AI | Founder OCG-Dubai | Helping organizations navigate the "AI Gold Rush"

[Genco Divrikli @genco_dvrkl]

@ForrestPKnight Yeah basically Codex is the engineer — structured, disciplined, builds clean. Opus is the inventor — takes risks, connects dots laterally, surprises you.
Different prompting because different cognitive wiring. And the smart move is chaining them — Opus to imagine, Codex to build.

[Genco Divrikli @genco_dvrkl]

@shiri_shh Nice move

[Genco Divrikli @genco_dvrkl]

@andyreed Haha, try multi agents😂

[Genco Divrikli @genco_dvrkl]

@TanerTekerX Thanks! 👍

[Genco Divrikli @genco_dvrkl]

@0xgaut 😂😂

[Genco Divrikli @genco_dvrkl]

@cgtwts 🤣🤣

[Genco Divrikli @genco_dvrkl]

@dev_maims Problem solving is the skill of this new age

[Genco Divrikli @genco_dvrkl]

@DonRenaldo57 @Dan_Jeffries1 True

[Genco Divrikli @genco_dvrkl]

@embirico Was having same experience initially but creativity , problem solving and desire to solve / adress issue is still in chatgpt standards

[Genco Divrikli @genco_dvrkl]

@godofprompt True

[Genco Divrikli @genco_dvrkl]

@metinaxu Yapay zeka muhendis olabilir

[Genco Divrikli @genco_dvrkl]

@ujjwalscript As long as user solves context accumulation and experienced in phased implementation with swe background i agree with the statement of Anthropic

[Genco Divrikli @genco_dvrkl]

@damnGruz Haha, then weekly limit hits 🤣

[Genco Divrikli @genco_dvrkl]

@tristanbob Claude Max or GLM Max (if you use ralph loop)

[Genco Divrikli @genco_dvrkl]

@TanerTekerX haha, sonu super gercekten 😂

[Genco Divrikli @genco_dvrkl]

https://t.co/ldbBNipUhY

[Lee Robinson @leerob]

Writing a blog post about research we're doing on long-running autonomous coding agents.

One example ran for *3 weeks* to build something of significant technical complexity.

It took a custom harness, and the best model was not what you might expect. Was surprising to me.

[Genco Divrikli @genco_dvrkl]

@chaseadams @TheAhmadOsman @Chaseadams, I issued a local ai setup guide recently , please have a check if that it addresses your inquiry. https://t.co/UckBjwyOZG

[Genco Divrikli @genco_dvrkl]

3/9 - THE HARDWARE TRUTH
Entry Level ($250-800)

Used RTX 3060 12GB ($200-350) - Best budget entry, runs 7B-13B models smoothly
RTX 4060 8GB ($280-320) - Runs 7B models well; 13B requires heavy quantization
Intel Arc B580 12GB ($249) - New budget king, 62 tok/s on 8B models
RTX 5070 12GB ($549 MSRP) - Blackwell architecture, great efficiency
Used M1/M2 MacBook 16GB ($500-800) - Surprisingly capable for small models
Mac Mini M4 16GB ($599 new) - Great starter Apple Silicon option

⚠️ Note: RTX 4060 has 8GB VRAM (not 12GB) - important limitation for larger models

[Genco Divrikli @genco_dvrkl]

@TheAhmadOsman Great point @Ahmad , how did you tackle memory issue ? and as far I know Minimax and GLM models in opencode is provided through Opencode Zen servers for free.