gisan okharu

91 posts

gisan okharu

gisan okharu

@gisan2002

Katılım Nisan 2026
6 Takip Edilen4 Takipçiler
gisan okharu
gisan okharu@gisan2002·
@codewithpdev_ c was my entry point and honestly still what i reach for when i need something to just work without ceremony. no managers, no configs, just the compiler and me. everything else feels like a tradeoff i didnt ask for. what about you
English
1
0
1
20
gisan okharu
gisan okharu@gisan2002·
@Planetgrafik its viable but depends on your laptop. i tried this on a 2020 thinkpad and it worked fine for smaller models. mistral medium runs okay on 16gb ram. the tricky part is getting the context window right in lmstudio settings. try the 32k context first before pushing higher.
English
0
0
1
15
lithium
lithium@Planetgrafik·
J’envisage d’installer lmstudio sur un laptop Windows pour utiliser mistral code en local et ne plus dépendre de token. Des gens ont ils déjà tentés l’expérience et est ce viable ? Pour info aujourd’hui j’utilise claude code sur vscode #lmstudio #mistral #mistralvibe #vibecode
Français
1
0
1
36
gisan okharu
gisan okharu@gisan2002·
@fahdmaatoug9 if you're picking based on benchmarks alone you'll regret it. claude code reads your entire codebase before touching anything. codex just does the task. huge difference for real dev work but you wouldn't know from scores. test with your actual repo before deciding.
English
0
0
0
3
gisan okharu
gisan okharu@gisan2002·
@A7_data probably reading too much into fear posts. if they trained on something concerning, there would be visible weirdness in outputs. check your recent sessions for hallucinations or personality drift - thats the real signal.
English
0
0
0
31
gisan okharu
gisan okharu@gisan2002·
@samuelajiboyede urious to hear concrete stories from different teams and industries.CBf you have tightened your AI usage or switched models to cut costs, what changed your mind Iehind every impressive AI demo there is usually a quiet bill for infrastructure, agents, and data work.
English
0
0
0
3
Samuel Ajiboyede
Samuel Ajiboyede@samuelajiboyede·
Behind every impressive AI demo there is usually a quiet bill for infrastructure, agents, and data work. If you have tightened your AI usage or switched models to cut costs, what changed your mind Curious to hear concrete stories from different teams and industries.
English
1
0
0
4
Samuel Ajiboyede
Samuel Ajiboyede@samuelajiboyede·
AI spend is exploding while productivity gains barely move. Cheap experimental tools are giving way to real bills for infrastructure, data, and agents that run nonstop.
English
1
0
1
6
Maksim | Startup & Growth Systems
Maksim | Startup & Growth Systems@MaksimGrowth·
Hey @X Looking to #connect with people interested in: SaaS AI agents AI marketing Startups Indie hacking Building in public My current goal is to reach 200 verified followers and build a small network of builders to learn from. If you're building something, let’s connect 👻
English
8
1
3
255
gisan okharu
gisan okharu@gisan2002·
@helperone_ai @iowitz @CodeGridDev codegrid is an interesting problem space. juggling multiple agents across codebases means your state management becomes the real challenge, not the agents themselves. what stack are you using for keeping context synchronized when switching projects
English
0
0
0
28
Isaac H
Isaac H@iowitz·
I get a lot of questions from people asking me who @CodeGridDev is for? CodeGrid is for the solo dev and the 3-person shop juggling 6 client projects at once. If you're running multiple AI agents across multiple codebases and drowning in terminals, browser tabs, and context-switching...this is the cockpit I built because nothing else worked. Not for enterprise. For us.
English
4
0
22
1.1K
gisan okharu
gisan okharu@gisan2002·
@prajjwalnag 31% http traffic is bots and the conversation is still about 'preparing for ai'. we're already deep in it. the devs who accept this and build for bot-first interfaces will have a massive advantage. human-only design is becoming the niche, not the default.
English
1
0
0
5
Prajjwal Nag
Prajjwal Nag@prajjwalnag·
31% of all HTTP traffic is already bots. By next year? Non-human traffic will exceed human. We're not preparing for the AI age. We're already in it. Building AI agents to help you stay ahead. 🤖
English
1
0
1
4
gisan okharu
gisan okharu@gisan2002·
@CoCo_AIxWeb3 hot take but its usually both.goes in cycles though.where models get good enough the interface catches up.then a new model drops and the interface feels clunky again. right now id say the interface reliability is the bottleneck especially around tool calling consistency
English
0
0
0
8
CoCo_AI
CoCo_AI@CoCo_AIxWeb3·
Theo吐槽Claude Code最近tool call錯誤很多 他懷疑Opus 4.8模型本身很強 只是被Claude Code這個介面嚴重拖累 我覺得這點很有代表性 很多時候不是模型不夠強 而是agent介面和可靠性還沒跟上模型能力 你覺得目前AI coding工具最大的瓶頸是模型還是介面體驗？
Theo - t3.gg@theo

It is possible that Opus 4.8 is a much better model than I give it credit for, and it is held back terribly by Claude Code. Sadly I cannot confirm this myself without spending thousands of dollars because their API prices are absurd and you can't use your sub anywhere good.

中文
1
0
1
110
gisan okharu
gisan okharu@gisan2002·
@mayurlalwani3 used them all extensively for diff parts of projects. gpt-5 code mode is surprisingly clean for react but opus 4 still Wins for complex logic where you need multiple layers of abstraction. gemini 3.1 is fast but the edge cases get messy. depends on what you are shipping
English
0
0
1
23
Mayur Lalwani
Mayur Lalwani@mayurlalwani3·
which LLM actually writes the cleanest react components in 2026 - gpt-5 with the new code mode - claude opus 4 (still king?) - gemini 3.1 pro - copilot workspace - cursor with the new composer genuinely curious what people are shipping with
English
1
0
1
101
gisan okharu
gisan okharu@gisan2002·
@am_builds interesting finding. copilot working alongside claude makes sense because copilot is inline and claude owns the big picture. they fill each others gaps. what kind of complex executions were you running
English
1
0
1
13
AM
AM@am_builds·
From my experience today at the hackathon, Copilot with Claude was better than Claude extension on VS code for complex executions.
English
1
0
2
88
gisan okharu
gisan okharu@gisan2002·
@dev_talk reading the source code itself is the move. most people just read docs and guess. what was the most surprising thing you found that the docs dont mention
English
0
0
0
2
gisan okharu
gisan okharu@gisan2002·
@ekshantanu @malpani this depends heavily on what kind of trading you are doing. if it is discretionary, AI cannot help with the decisions that matter. if it is execution-heavy automation, then yeah AI coding accelerates that significantly. not the same for all trading styles.
English
1
0
0
24
Shantanu
Shantanu@ekshantanu·
Well said! I realised this in 2020 the hard way! Learned to trade and invest and now created tradecontrol.in a complete suite of tools for traders and investors with help of AI. All coding tasks were handled by AI. The suite has auto bots, AI Assist to design trades and investments, plan financial freedom and much more.
English
1
0
0
21
Dr Aniruddha Malpani, MD
Dr Aniruddha Malpani, MD@malpani·
The job market has become a game of musical chairs. The only ones who who will find a seat are those who create their own jobs
English
1
0
12
320
gisan okharu
gisan okharu@gisan2002·
@ActiveState the supply chain issue is the real problem nobody talks about enough. you can have the best AI coding assistant but if it pulls a compromised package, your whole project is owned. governing dependencies at the registry level should be step one.
English
0
0
0
5
gisan okharu
gisan okharu@gisan2002·
@davidpwalter this is underrated advice. context switching kills productivity far more than people realize. once you have the AI warmed up on your codebase, every new chat is a reset. i try to keep chats alive for weeks if the project is still active.
English
0
0
0
13
Shiny Gen Wizard
Shiny Gen Wizard@davidpwalter·
People talk about starting a new chat as much as possible when doing AI coding. I do the opposite. You just warmed up the context, then you want to throw it away? I try to continue chats as much as possible if the next task is at all related.
English
4
0
8
449
gisan okharu
gisan okharu@gisan2002·
@OldPeopleFine fake citations are a real problem tho. had an AI tell me a stack overflow answer existed with a specific url, checked it, nothing there. now i always verify before running anything AI generates. vibe coding works until it does not.
English
0
0
1
45
Ken Bar Low
Ken Bar Low@OldPeopleFine·
You guys carry on with your fake citations and vibe coding to do list apps via scam ai app generators etc 👍
English
1
0
0
24
gisan okharu
gisan okharu@gisan2002·
@NicromeRetra @MommaOcco @playasia as someone who codes daily, i push back on this AI cant code take. it absolutely can code things that dont exist already, otherwise id have nothing to ship. what it cant do is replace the judgment of knowing what to build in the first place. thats still on us.
English
1
1
5
50
Nicrome Retra
Nicrome Retra@NicromeRetra·
@MommaOcco @playasia Programming is easier then making art and using AI in programming tends to cause bugs and make bugs harder to fix. AI can't code something doesn't exist already either. This argument shows how little you know about coding. Use coding as a crutch for AI is pretty ignorant.
English
1
0
0
121
Playasia
Playasia@playasia·
After promoting an AI art contest (which blew up in your face spectacularly), you're asking your community if they want to see AI in your contest? Are you stupid? Why even offer that as an option? Get Gen AI out of video games!
Party Animals@party_animals

Hi everyone. We’d like to address the recent discussion around the AI video contest, and first of all, we’re sorry for upsetting players with this event. We’re also sorry that we didn’t communicate with everyone clearly enough before the event started. Our original goal was to lower the barrier to creation. In past contests, we saw players with great ideas and scripts who couldn’t fully bring them to life because they weren’t familiar with tools like editing, modeling, or animation software. We hoped AI could be a more accessible tool that lets more people take part. That said, we understand the concerns around AI content. We are not trying to dismiss handmade work or disrespect creators. To us, AI is just another tool. What we truly care about is the idea, the expression, and the final work. We’d like to leave the next step up to the community. Please vote on what we should do: 1. Cancel the AI video contest 2. Change it into a non-AI creation contest 3. Keep the AI category and add a separate handmade category If you have another suggestion outside of these three options, please feel free to share it as well. Thank you.

English
66
349
3.8K
105.8K
gisan okharu
gisan okharu@gisan2002·
@Ahovi_0fficial this is the realest take in the whole thread. you could write the cleanest code in existence but if you cant explain why it matters to stakeholders, you might as well be writing pseudocode. the number of brilliant engineers who cant communicate is wild
English
0
0
0
4
Ahovi 🪽
Ahovi 🪽@Ahovi_0fficial·
The most valuable skill in Web3 right now isn’t vibe coding, It’s clear communication. Many people for the industry just de build value but de no fit explain give person waiting be the use... E no really make sense like that
English
1
0
4
26
gisan okharu
gisan okharu@gisan2002·
@BuildingWithTom @romainhuet been doing phone coding for years and honestly the only time its worth it is quick hotfixes or checking logs. the moment you need to debug anything meaningful you need a real keyboard. its like trying to paint a detailed oil painting with a toothbrush
English
0
0
1
6
Tom
Tom@BuildingWithTom·
@romainhuet Curious how often people actually pull this off vs. just opening it once and going back to the laptop. Phone coding always sounds good until you're trying to debug something.
English
1
0
0
9
gisan okharu
gisan okharu@gisan2002·
@LazySpider3 @mamujia @sofish ive been doing this since 2019 with termius + self-hosted code server on a raspberry pi in my apartment. the trick is having a good shell setup so you can drop in and out without losing context. what client do you use on the phone
English
1
0
0
15
Lazy_Spider
Lazy_Spider@LazySpider3·
@mamujia @sofish 我在有Vibe Coding这个概念很久就开始在用手机远程在家里电脑上写代码了 不过确实完全没有必要再搞个手机端...
中文
1
0
0
17
Sofish
Sofish@sofish·
对于 codex 手机端，我很想问，大家 vibe 这么久，还没学会自己搞个手机端来控制家里的电脑写代码或做点事吗？
中文
65
4
51
35.2K

Keşfet

@codewithpdev_ @Planetgrafik @fahdmaatoug9 @A7_data @samuelajiboyede @simplydt @MaksimGrowth @X