iiinova — where your digital self keeps evolving

𝕏 @x.laoda.ge.ap.brid.gy · 06/05/2026

Minimal illustration of [SUBJECT/OBJECT], clean thin lines with selective vibrant color accents, lots of negative space, refined composition, modern editorial style, elegant and simple. Prompt不错，风格很喜欢！ x.com/i/status/2049400614195794230

000 0

𝕏 @x.laoda.ge.ap.brid.gy · 22/01/2026

Gemini CLI is very stable these days.

000 0

𝕏 @x.laoda.ge.ap.brid.gy · 05/01/2026

😅 😂 We've got collateral damage!

000 0

𝕏 @x.laoda.ge.ap.brid.gy · 25/11/2025

Haha, Anthropic's claude limit is pushing users to use ChatGPT & Gemini. Never seen a stupid company like Anthropic before

000 0

𝕏 @x.laoda.ge.ap.brid.gy · 20/11/2025

The Generative UI of Gemini 3 is very powerful. A simple demo.

010 1

𝕏 @x.laoda.ge.ap.brid.gy · 15/11/2025

Installed adguard home on VPS.

000 0

𝕏 @x.laoda.ge.ap.brid.gy · 09/11/2025

Power consumption of my homelab server over the past 6 months. It's a Dell T5810 with a 2696v4 CPU & 256G of RAM & 1TB SATA SSD & 18TB HDD #homelab #selfhost

021 2

𝕏 @x.laoda.ge.ap.brid.gy · 16/10/2025

almost forget I need to replace my CPU. It took about 30 mins to finish the job.

000 0

𝕏 @x.laoda.ge.ap.brid.gy · 01/08/2025

Got a new beast! Still trying to figure out what's going on here! #malahit #malahitdsp2 #radio

000 0

𝕏 @x.laoda.ge.ap.brid.gy · 08/07/2025

Claude Code is overloaded.

000 0

𝕏 @x.laoda.ge.ap.brid.gy · 23/06/2025

第三个vibe coding项目差不多完结了，260+的commits，搞得很累，还是claude比gemini要好一些，很多复杂的功能居然一遍能过。关于安全加固，找了个图片，让它按照这个修复就行了。性能优化就靠纯Prompts了。

AI Vibe Coding Security Check
from: https://gist.github.com/mattppal/5c01ef4447e94515a03314db1ef2403e

000 0

𝕏 @x.laoda.ge.ap.brid.gy · 22/06/2025

Dunno if it's accurate

000 0

𝕏 @x.laoda.ge.ap.brid.gy · 19/06/2025

尽管acli rovodev每天免费的2000万token，但是体验不好，经常自己中断，中断后还得重新描述问题，好像没有memory bank，但是有时候它又在搜索我之前用roocode生成memory bank。总之用的很不舒服。Gemini Pro 2.5写代码能力相比claude还是差一些，尤其是最新的0605版本，最失败的一个版本。所以，又买了claude pro，结合claude code，用的很舒服。具体有多少token，也说不清楚，但是右下角有提示就很舒服。一般用量快到了，也就该休息了。上班摸鱼用pro足够了。 […] [Original post on laoda.ge]

010 1

𝕏 @x.laoda.ge.ap.brid.gy · 31/05/2025

😅 Links on the bottom are totally hallucinations.

000 0

𝕏 @x.laoda.ge.ap.brid.gy · 31/05/2025

Google quietly released an Android app named Google AI Edge Gallery which allows users run LLMs offline on their phone. github.com/google-ai-edge/gallery

100 1

𝕏 @x.laoda.ge.ap.brid.gy · 27/05/2025

This is 911, what's is your emergency?

100 1

𝕏 @x.laoda.ge.ap.brid.gy · 24/05/2025

Veo3 is available for Google AI Pro subscribers now. #veo3 #gemini

000 0

𝕏 @x.laoda.ge.ap.brid.gy · 23/05/2025

Anthropic's new AI model turns to blackmail when engineers try to take it offline | TechCrunch techcrunch.com/2025/05/22/anthropic…

000 0

𝕏 @x.laoda.ge.ap.brid.gy · 01/05/2025

Nvidia CEO Jensen Huang talks about the need for the US to change the rules for exporting AI technology to the rest of the world.

000 0

𝕏 @x.laoda.ge.ap.brid.gy · 30/04/2025

Meta AI is not that intelligent

000 0

𝕏 @x.laoda.ge.ap.brid.gy · 21/04/2025

Pope died 7 hours after this.

Pope died 7 hours after he met JD Vampire

000 0

𝕏 @x.laoda.ge.ap.brid.gy · 16/04/2025

Everyone can now use Veo2 for free in AI Studio to generate videos. #google #veo2 #aistudio

010 1

𝕏 @x.laoda.ge.ap.brid.gy · 11/04/2025

Images generated by ChatGPT 4o.

000 0

𝕏 @x.laoda.ge.ap.brid.gy · 20/01/2025

To support the research community, we have open-sourced DeepSeek-R1-Zero, DeepSeek-R1, and six dense models distilled from DeepSeek-R1 based on Llama and Qwen. DeepSeek-R1-Distill-Qwen-32B outperforms OpenAI-o1-mini across various benchmarks, achieving new […] [Original post on laoda.ge]

001 1

𝕏 @x.laoda.ge.ap.brid.gy · 11/01/2025

Very impressive image recognition model. VRAM requirements: <3GB github.com/vikhyat/moondream #LLM #selfhost #moondream

014 4

𝕏 @x.laoda.ge.ap.brid.gy · 25/12/2024

Got both linkwarden & hoarder installed on my homelab server. But I'm not sure which I'll ultimately keep. Both are excellent. github.com/linkwarden/linkwarden github.com/hoarder-app/hoarder #Homelab #selfhost #bookmark #linkwarden #hoarder

011 1

𝕏 @x.laoda.ge.ap.brid.gy · 20/12/2024

How come no one did this before?

000 0

𝕏 @x.laoda.ge.ap.brid.gy · 20/12/2024

"Mangione was charged in New York with first degree murder in furtherance of terrorism. Under New York law, the terrorism charge can be brought if the act "intended to intimidate or coerce a civilian population, influence the policies of a unit of government by […] [Original post on laoda.ge]

000 0

𝕏 @x.laoda.ge.ap.brid.gy · 09/12/2024

Sora has been officially launched by OpenAI, but new user registration has been temporarily disabled. #sora #OpenAI

000 0

𝕏 @x.laoda.ge.ap.brid.gy · 06/12/2024

Sam Altman's statement about ChatGPT Pro users having unlimited access to o1, 4o and o1-pro model is not accurate. o1 is limited to 100 conversations per day, while 4o has a limit of 600. The usage limit of o1-Pro is unknown yet, but it's clearly much lower […] [Original post on laoda.ge]

000 0

𝕏 @x.laoda.ge.ap.brid.gy · 05/12/2024

AI Hallucination is inevitable. What we can do is try our best to minimize it 不by writing more sophisticated promots as end users. AI产生幻觉很难避免，作为终端用户，我们能做的就是写出更复杂详细的prompts，而这个是我最讨厌的。 aiup.date/podcast/2024/12/05/AI%20H…

000 0

𝕏 @x.laoda.ge.ap.brid.gy · 04/12/2024

#OpenAI got lots of new things to talk about.

000 0

𝕏 @x.laoda.ge.ap.brid.gy · 30/11/2024

有了Claude的mcp之后，我已经完全的废了！！ #Claude #MCP

Claude is enhancing my local app using mcp!

000 0

𝕏 @x.laoda.ge.ap.brid.gy · 24/11/2024

MagicQuill is an intelligent and interactive system achieving precise image editing. Yes! You can selfhost it! github.com/magic-quill/MagicQuill #selfhost

001 1

𝕏 @x.laoda.ge.ap.brid.gy · 18/11/2024

CodeImage is the newest tool to help developers to create beautiful screenshots of their code, providing several features to speed up the process to post in social media. You can selfhost it very easily. github.com/riccardoperra/codeimage #CodeImage #selfhost

111 1

𝕏 @x.laoda.ge.ap.brid.gy · 16/11/2024

This thing will be all over the battlefield soon. #robot #robotdog

000 0

𝕏 @x.laoda.ge.ap.brid.gy · 14/11/2024

This is why I have to double the RAM of my homelab server.

000 0

𝕏 @x.laoda.ge.ap.brid.gy · 12/11/2024

Successfully upgraded RAM on my homelab server, strangely enough, the server did't boot up properly on the first restart. But when I plugged in a keyboard and monitor to try to troubleshoot the issue, it ended up booting up just fine. It's not a bad thing to always have enough RAM.

successfully upgraded homelab server RAM

000 0

𝕏 @x.laoda.ge.ap.brid.gy · 12/11/2024

Is Role Prompting Effective? Vanilla prompts (like you are an expert in some field) produce limited or no positive effects. In some cases they may lead to positive outcomes.😲 But expert (complicated) pormpts can do a much better job. Blog […] [Original post on laoda.ge]

Idiot prompt is better than genius prompt

000 0

𝕏 @x.laoda.ge.ap.brid.gy · 12/11/2024

Alibaba has released Qwen2.5-Coder-32B-instruct, claiming it to be the most powerful coding LLM to date. qwenlm.github.io/blog/qwen2.5-coder…

000 0