# ![@0xSero Avatar](https://lunarcrush.com/gi/w:26/cr:twitter::1341331548566654977.png) @0xSero 0xSero

0xSero posts on X about ai, if you, more than, claude code the most. They currently have [------] followers and [---] posts still getting attention that total [------] engagements in the last [--] hours.

### Engagements: [------] [#](/creator/twitter::1341331548566654977/interactions)
![Engagements Line Chart](https://lunarcrush.com/gi/w:600/cr:twitter::1341331548566654977/c:line/m:interactions.svg)

- [--] Week [-------] -73%
- [--] Month [---------] +72%
- [--] Months [---------] +19,335%
- [--] Year [---------] +1,007%

### Mentions: [--] [#](/creator/twitter::1341331548566654977/posts_active)
![Mentions Line Chart](https://lunarcrush.com/gi/w:600/cr:twitter::1341331548566654977/c:line/m:posts_active.svg)

- [--] Week [---] -18%
- [--] Month [---] +86%
- [--] Months [---] +1,392%
- [--] Year [---] +689%

### Followers: [------] [#](/creator/twitter::1341331548566654977/followers)
![Followers Line Chart](https://lunarcrush.com/gi/w:600/cr:twitter::1341331548566654977/c:line/m:followers.svg)

- [--] Week [------] +6%
- [--] Month [------] +47%
- [--] Months [------] +790%
- [--] Year [------] +1,807%

### CreatorRank: [-------] [#](/creator/twitter::1341331548566654977/influencer_rank)
![CreatorRank Line Chart](https://lunarcrush.com/gi/w:600/cr:twitter::1341331548566654977/c:line/m:influencer_rank.svg)

### Social Influence

**Social category influence**
[technology brands](/list/technology-brands)  20% [stocks](/list/stocks)  3.33% [social networks](/list/social-networks)  2.5% [products](/list/products)  1.67% [currencies](/list/currencies)  0.83% [luxury brands](/list/luxury-brands)  0.83% [automotive brands](/list/automotive-brands)  0.83% [finance](/list/finance)  0.83% [cryptocurrencies](/list/cryptocurrencies)  0.83%

**Social topic influence**
[ai](/topic/ai) 6.67%, [if you](/topic/if-you) 5.83%, [more than](/topic/more-than) 5%, [claude code](/topic/claude-code) #2649, [anthropic](/topic/anthropic) #3624, [llm](/topic/llm) #514, [the most](/topic/the-most) #3063, [up to](/topic/up-to) 2.5%, [in the](/topic/in-the) 2.5%, [crypto](/topic/crypto) #5074

**Top accounts mentioned or mentioned by**
[@itsjustmarky](/creator/undefined) [@alexfinn](/creator/undefined) [@kr0der](/creator/undefined) [@zenmagnets](/creator/undefined) [@bitdeep_](/creator/undefined) [@alexcar12](/creator/undefined) [@willccbb](/creator/undefined) [@marcohmann](/creator/undefined) [@hlitoroy](/creator/undefined) [@royjossfolk](/creator/undefined) [@interpreter_ai](/creator/undefined) [@007killpop](/creator/undefined) [@tomcocobrico](/creator/undefined) [@hotaisle](/creator/undefined) [@tcgsealed](/creator/undefined) [@cacus](/creator/undefined) [@jason](/creator/undefined) [@openclaw](/creator/undefined) [@uttkarsh26](/creator/undefined) [@sudoingx](/creator/undefined)

**Top assets mentioned**
[Ethereum (ETH)](/topic/ethereum)
### Top Social Posts
Top posts by engagements in the last [--] hours

"Heaven is home"  
[X Link](https://x.com/0xSero/status/2021997746413240457)  2026-02-12T17:20Z 10.8K followers, [----] engagements


"I have been using 5-10 Claude Opus instances 12-24 hours a day on the 200$ a month plan. I am not one to shill but the value on this is incredible for now. I feel Anthropic has been getting much better at listening"  
[X Link](https://x.com/0xSero/status/1996190510076273119)  2025-12-03T12:11Z 10.6K followers, 69.2K engagements


"My GPUs sound like a jet engine. I love the sound mining javascript like a real cryptobro"  
[X Link](https://x.com/0xSero/status/2012913561626976549)  2026-01-18T15:42Z 10.1K followers, [----] engagements


"Today I am investigating the profitability of the Anthropic API I will use public information and my personal usage to answer the question: What is the real cost of a token Video on YouTube today"  
[X Link](https://x.com/0xSero/status/2014626036634308742)  2026-01-23T09:07Z 10.7K followers, [----] engagements


"Are you interested in running LLMs locally Here is a full breakdown of all the hardware options the costs what models you can run and general advice I wish I knew. I showed how I use my local models as well (: enjoy. https://youtu.be/Adliwsf2oPE https://youtu.be/Adliwsf2oPE"  
[X Link](https://x.com/0xSero/status/2015839443354030375)  2026-01-26T17:29Z 10.1K followers, 35.7K engagements


"2 videos coming today: [--]. Kimi-K2.5: I have been paying for their basic plan since they added Nano Banana slides and absolutely love it. This is a huge upgrade. [--]. Local LLM tier list: both will be an exploratory brain dump. https://www.youtube.com/channel/UCeNLqAes-kUMjWksPA6d-BA/ https://www.youtube.com/channel/UCeNLqAes-kUMjWksPA6d-BA/"  
[X Link](https://x.com/0xSero/status/2016102610772599005)  2026-01-27T10:54Z 10.1K followers, 10.8K engagements


"Kimi IS COOKING holy mackerel this is way better than anything I can get out of opus or GPT Has some bugs. but looks soooo unique and well into my brand for [--] shot I can't complain"  
[X Link](https://x.com/0xSero/status/2016110996071354666)  2026-01-27T11:28Z 10.3K followers, 240K engagements


"Kimi-K2.5 running in Pi Opencode and Droid. Let's goooo"  
[X Link](https://x.com/0xSero/status/2016120872273096712)  2026-01-27T12:07Z 10.1K followers, 11.2K engagements


"I just reviewed Kimi-K2.5 it is genuinely the most impressive open-weight model I have tried it's great at: - Coding - Vision - Computer use - Web design - Games - Chatting - Slides I think we finally broke the AI UI design slop curse As promised: https://youtu.be/hVtofKryTqA https://youtu.be/hVtofKryTqA"  
[X Link](https://x.com/0xSero/status/2016159932152766797)  2026-01-27T14:42Z 10.2K followers, 18.9K engagements


"I think I found the [--] besides a few UX issues this model is producing EXTREMELY unique websites with very little steering or input. I think this will be my new daily driver for frontend development and computer use. Cerebras needs to help them speed their inference up"  
[X Link](https://x.com/0xSero/status/2016161041625805003)  2026-01-27T14:47Z 10.1K followers, 14.6K engagements


"Kimi-K2.5 before and after for vllm-studio Before: After: https://twitter.com/i/web/status/2016265050269290564 https://twitter.com/i/web/status/2016265050269290564"  
[X Link](https://x.com/0xSero/status/2016265050269290564)  2026-01-27T21:40Z 10.1K followers, [----] engagements


"You can spam it pretty hard at 40$ p/m"  
[X Link](https://x.com/0xSero/status/2016667756691673541)  2026-01-29T00:20Z 10.2K followers, 14.7K engagements


"The only option is to win. If you surround yourself with people who are complacent who lack agency and the ability to tell the truth you will lose so hard. Regardless of whether a bubble pops or we are in the middle of a singularity there is no choice to be weak or slow"  
[X Link](https://x.com/0xSero/status/2016870076410900793)  2026-01-29T13:44Z 10.2K followers, [----] engagements


"I am rebuilding my website with Kimi-K2.5 God damn man I love the vibes it's been so fun working on frontend stuff"  
[X Link](https://x.com/0xSero/status/2016930260651053166)  2026-01-29T17:43Z 10.2K followers, [----] engagements


"Before & After Kimi-K2.5"  
[X Link](https://x.com/0xSero/status/2017019353560912262)  2026-01-29T23:37Z 10.3K followers, [----] engagements


"This is the killer use case for Cerebras. Write a change you want made or speak it into whisper and have it instantly fix it all"  
[X Link](https://x.com/0xSero/status/2017356148047827118)  2026-01-30T21:56Z 10.2K followers, [----] engagements


"I just nuked [--] different apps of my Macbook and took all my old code cleared the node_modules and moved them into an archive. My device has never been messier but finally tending to the land"  
[X Link](https://x.com/0xSero/status/2017717054686495036)  2026-01-31T21:50Z 10.3K followers, [----] engagements


"gpt-oss-120b is good if they open source something bigger and better oooof"  
[X Link](https://x.com/0xSero/status/2017748590525816883)  2026-01-31T23:55Z 10.1K followers, [----] engagements


"I am ripping out all my ai-sdk/agent/filesystem slopgic and replacing it with Pi-ai a copy of pods and agent-core. Its very well written. Wish I dug deeper earlier save myself the hassle. https://github.com/badlogic/pi-mono https://github.com/badlogic/pi-mono"  
[X Link](https://x.com/0xSero/status/2017944775110910107)  2026-02-01T12:55Z 10.2K followers, 32.9K engagements


"Oyoyoy sir: [--]. GLM-4.7-flash [--]. GLM-4.5-Air [--]. GPT-OSS-120B Run them with full context on vLLM if you can use fp8 kv cache If you have 192gb that opens the door for GLM-4.7-Reap and MiniMax-M2.1 this is where it truly starts getting Sonnet-like. Run these on whatever software and hook it up to a cloudflared to serve. https://twitter.com/i/web/status/2018019370438385736 https://twitter.com/i/web/status/2018019370438385736"  
[X Link](https://x.com/0xSero/status/2018019370438385736)  2026-02-01T17:51Z 10.2K followers, [----] engagements


"For those with 24-96gb of vram/unified memory. I highly recommend trying out GLM-4.7-Flash. It is the best model up to 100B params I showed how I use it and what it can do. GLM-4.7-Flash - Claude Code at home https://youtu.be/_SDyaPYmIxU https://youtu.be/_SDyaPYmIxU"  
[X Link](https://x.com/0xSero/status/2018027097084588503)  2026-02-01T18:22Z 10.4K followers, 33.2K engagements


"@tofikyeah Why do you have your private key running on device bro. I make burners to do stuff all the time but 70k belongs on a hardware wallet"  
[X Link](https://x.com/0xSero/status/2018056541400666149)  2026-02-01T20:19Z 10.4K followers, [----] engagements


"As someone who really wants this to work its going to be a slopfest completely dysfunctional Every subagent will increase the distance between what we want and what we will get increase the likelihood of issues decrease traceability etc. Really hope Im wrong 🚨 Claude Code LEAKS It can now spawn specialized agents that work on tasks like teammates Each gets a detailed brief and builds autonomously Runs in background while you keep chatting Multiple agents work in parallel on different parts Basically a dev team in your 🚨 Claude Code LEAKS It can now spawn specialized agents that work on"  
[X Link](https://x.com/0xSero/status/2018251350782435672)  2026-02-02T09:13Z 10.1K followers, [----] engagements


"@nummanali Yes its telegram. I used discord but too much complexity around context"  
[X Link](https://x.com/0xSero/status/2018261159816679692)  2026-02-02T09:52Z 10.3K followers, [---] engagements


"Let me make your workflows a bit easier [--]. Get a GPT sub (you should do this if you code esp) [--]. Install the Codex app [--]. Setup your project root directory [--]. Setup an automation 24h Interval [--]. "Analyze all my git history changes over the week give me recommendations and advice" The goal here is to get some visibility into your work performance this is model agnostic too. - How much code did you ship - What projects did you commit to the most - What bad coding habits can you improve - What patterns did you apply in [--] repo that you can apply in another I have been doing this manually for"  
[X Link](https://x.com/0xSero/status/2018435827529826365)  2026-02-02T21:26Z 10.5K followers, 18.6K engagements


"Ray and I go deep into local AI hardware and various usecases for private AI. This was a great really enjoyed this conversation especially since I've been listening to his podcast for so long. https://youtu.be/dHWD6DIiR-s https://youtu.be/dHWD6DIiR-s"  
[X Link](https://x.com/0xSero/status/2018466754834706771)  2026-02-02T23:29Z 10.2K followers, [----] engagements


"More pewter more pewter"  
[X Link](https://x.com/0xSero/status/2018624148436914345)  2026-02-03T09:54Z 10.7K followers, [----] engagements


"Mac Mini just arrived Im very excited for this. my 3rd Mac lol"  
[X Link](https://x.com/0xSero/status/2018653356559052910)  2026-02-03T11:50Z 10.3K followers, 15K engagements


"Lets go Qwen-next-coder"  
[X Link](https://x.com/0xSero/status/2018800642681352636)  2026-02-03T21:35Z 10.1K followers, [----] engagements


"This was an idea thought up by openclaw with Kimi-K2.5 and executed by Opus [---] A content creation machine battles of the clankers. The idea: "put [--] models against each other and have them produce visual output using ascii" Using Clawdbot peekaboo and the heartbeat feature every [--] hours it'll make a prompt to put [--] models against eachother record it on the macmini then send it to me on telegram. Left: Opus Right: Codex [---] high https://twitter.com/i/web/status/2018819536233939449 https://twitter.com/i/web/status/2018819536233939449"  
[X Link](https://x.com/0xSero/status/2018819536233939449)  2026-02-03T22:50Z 10.3K followers, [----] engagements


"Droid Pi and Opencode I will shill to my last breath. Best tools in the space. Droid now supports plugins. Plugins bundle skills commands and agents into packages you can install and share. Every team works differently. Plugins let you shape Droid to fit your workflow then share that setup with others on your team. https://t.co/hpifApQQep Droid now supports plugins. Plugins bundle skills commands and agents into packages you can install and share. Every team works differently. Plugins let you shape Droid to fit your workflow then share that setup with others on your team."  
[X Link](https://x.com/0xSero/status/2018836133187351028)  2026-02-03T23:56Z 10.1K followers, 20K engagements


"@AlexFinn 10$ - Minimax 6$ - GLM 40$ - Kimi 40$ - 2x GPT subs 100$ - Claude 0$ - GLM-4.7 0$ - openrouter/free (routes you to best free models) Literally more than you can use in a month regardless how much you spam them"  
[X Link](https://x.com/0xSero/status/2018837188860096921)  2026-02-04T00:01Z 10.3K followers, [----] engagements


"@karpathy Karpathy relentlessly optimising NanoGPT over and over until we get AGI 😂"  
[X Link](https://x.com/0xSero/status/2018960768121639406)  2026-02-04T08:12Z 10.1K followers, [----] engagements


"Claude Code as a harness is good enough for 95% of people. Its got a ton of features more than needed its got a great model behind it too. Not knocking on Anthropic. However: [--]. The context problem: Claude code crams so much trash into the model (60k tokens by default) so you only get 140k tokens out of the model. [--]. It has severe performance issues running very long sessions using subagents etc. chews through RAM and CPU time more than any other harness. [--]. Opus is amazing but its far from the best at everything codex is better for programming planning debugging and architecture."  
[X Link](https://x.com/0xSero/status/2018962960190738563)  2026-02-04T08:20Z 10.3K followers, [---] engagements


"Supply will catch up to demand over the next few years especially as we get more tailor made chips for inference. Similar to ASIC Miners I remember when people were using GPUs to mine crypto now no one does that. Also still the price of electronics/tech over [--] years has gone down tremendously. A spike for a few years isnt the end of the world"  
[X Link](https://x.com/0xSero/status/2018967482011107384)  2026-02-04T08:38Z 10.2K followers, [----] engagements


"Let's see how oh my opencode does with Kimi. Scrape [----] data points from the web (:"  
[X Link](https://x.com/0xSero/status/2018985639782596964)  2026-02-04T09:51Z 10.2K followers, [----] engagements


"@HotAisle I didnt know you did this What were the economics of it like What GPUs were u using"  
[X Link](https://x.com/0xSero/status/2019163141918056950)  2026-02-04T21:36Z 10.1K followers, [---] engagements


"Let me make your life better. Do you want to use the GUI of any device from your laptop/macbook pc SSH is great but sometimes you want a GUI. [--]. Take the devices you want to hook up [--]. Install tailscale (free) on all of them [--]. Install NoMachine on both the controlling laptop and the remote devices [--]. Run NoMachine it'll expose port [----] on your local network [--]. In NoMachine on the controller just put your login details [--]. Now my macbook can control the GUI of my linux AI rig my mac mini my old laptop/mac mini etc. Performs well enough and is relatively secure if you use tailscale"  
[X Link](https://x.com/0xSero/status/2019187296944165293)  2026-02-04T23:12Z 10.3K followers, [----] engagements


"@cosmic_spec MiniMax-M2.1 + GLM"  
[X Link](https://x.com/0xSero/status/2019298043225330042)  2026-02-05T06:32Z 10.2K followers, [--] engagements


"Still thinking about this. I was a bit anxious going in I am very confident in what I bring to the table and genuinely think I am on the more cracked end of the agentic coding niche. What helped a lot was writing pseudo code which took me [--] minutes to solve in natural language. Translating that to code was tough I was struggling with basic syntax Id used Typescript Map() for years. I was trying to map.add() which is not a thing lol but the interviewer allowed me to look up the Syntax which was very nice. The main problem I feel bad about was how I solved the problem not the data structures I"  
[X Link](https://x.com/0xSero/status/2019329319282024850)  2026-02-05T08:36Z 10.2K followers, [----] engagements


"@Alex_Car12 @willccbb [--]. You dont need to use the ACP panel just use the terminal [--]. You can add your provider and api keys DIRECTLY into the Zed agent which gives you the full history [--]. The window tab feature is the main reason this is my daily driver now. I can have [--] projects in [--] window"  
[X Link](https://x.com/0xSero/status/2019336953921995231)  2026-02-05T09:07Z 10.2K followers, [--] engagements


"@Alex_Car12 @willccbb Pin the terminal to the right side and itll be just like cursor :p"  
[X Link](https://x.com/0xSero/status/2019337043994939595)  2026-02-05T09:07Z 10.1K followers, [--] engagements


"@Alex_Car12 @willccbb If you want there's a repo on my github for the PI acp it has session management built into it"  
[X Link](https://x.com/0xSero/status/2019359419260432449)  2026-02-05T10:36Z 10.2K followers, [--] engagements


"Been using GLM-4.7-Flash for my Clawdbot it works just fine it's keeping track of the news for me creating PDFs on trends every [--] hours It's managing it's browser for a model running on so little hardware I wasn't expecting it to handle the types of tasks that it does https://twitter.com/i/web/status/2019375993538916673 https://twitter.com/i/web/status/2019375993538916673"  
[X Link](https://x.com/0xSero/status/2019375993538916673)  2026-02-05T11:42Z 10.5K followers, [----] engagements


"@port_dev This could run on 96GB vram with full context even 48gb realistically. For me: - 192GB vram 8x3090s"  
[X Link](https://x.com/0xSero/status/2019416396502036621)  2026-02-05T14:22Z 10.2K followers, [---] engagements


"Is vllm-studio beautiful or what (:"  
[X Link](https://x.com/0xSero/status/2019442229245256181)  2026-02-05T16:05Z 10.2K followers, [----] engagements


"One of the better and cheaper options for openclaw if you have more than 128gb mixed memory you can also run it locally https://t.co/K9YAoLzEKA https://t.co/K9YAoLzEKA"  
[X Link](https://x.com/0xSero/status/2019671842018197527)  2026-02-06T07:17Z 10.3K followers, [----] engagements


"@TCGsealed @Alibaba_Qwen I love Qwen but its fair to criticise"  
[X Link](https://x.com/0xSero/status/2019772560297042425)  2026-02-06T13:57Z 10.3K followers, [---] engagements


"This bad boy came in so clutch today. The internet provider where I live lost service for [--] hours today that didnt slow me down [--] bit. Now heres the problem I need more GPUs I need to get Kimi running locally. Thatd be GG"  
[X Link](https://x.com/0xSero/status/2019857589341663465)  2026-02-06T19:35Z 10.3K followers, [----] engagements


"@ExileAI_0 GLM-4.7-Flash with MLX or vLLM"  
[X Link](https://x.com/0xSero/status/2019868228873564253)  2026-02-06T20:18Z 10.2K followers, [---] engagements


"@marcohmann Even on a 50% reap with [--] bit quant itd still be 250GB vram also I am sure vision would break significantly"  
[X Link](https://x.com/0xSero/status/2019868425242505668)  2026-02-06T20:18Z 10.3K followers, [---] engagements


"@cacus Its amazing but the m3 speed isnt fast enough to warrant 25k usd. If the m5 ultra is even 2x as fast I will cop"  
[X Link](https://x.com/0xSero/status/2019882843636396449)  2026-02-06T21:16Z 10.3K followers, [---] engagements


"I might be dumber than 90% of my peers but I will work till 1am to make up for it"  
[X Link](https://x.com/0xSero/status/2019920645526798625)  2026-02-06T23:46Z 10.2K followers, [----] engagements


"ExoLabs will win so hard you have no idea. Ill be honest I have [--] mac minis. [--] more clusters like this one. Why @jason thought my argument for local AI would be cost but its much more than that. AI is becoming an extension of your brain an exocortex. @openclaw is a huge leap towards that. It knows everything you https://t.co/o6Y3psSTYB Ill be honest I have [--] mac minis. [--] more clusters like this one. Why @jason thought my argument for local AI would be cost but its much more than that. AI is becoming an extension of your brain an exocortex. @openclaw is a huge leap towards that. It knows"  
[X Link](https://x.com/0xSero/status/2020062419431698787)  2026-02-07T09:09Z 10.5K followers, 15.6K engagements


"@AlexFinn @alexocheema @Jason @openclaw - Whisper - Qwen-image-edit - TTS"  
[X Link](https://x.com/0xSero/status/2020062605797179763)  2026-02-07T09:10Z 10.3K followers, [---] engagements


"I wonder why Claude loves shrimp and lobsters so much Make the first contact. https://t.co/CzBvHL71ss Make the first contact. https://t.co/CzBvHL71ss"  
[X Link](https://x.com/0xSero/status/2020072114234220633)  2026-02-07T09:48Z 10.3K followers, [----] engagements


"@alex_bit_ Yes thats the idea"  
[X Link](https://x.com/0xSero/status/2020112139738886384)  2026-02-07T12:27Z 10.3K followers, [--] engagements


"@kr0der Yes"  
[X Link](https://x.com/0xSero/status/2020126123183542513)  2026-02-07T13:22Z 10.3K followers, [---] engagements


"@uttkarsh26 @hlitoroy Not really"  
[X Link](https://x.com/0xSero/status/2020134441037742088)  2026-02-07T13:55Z 10.3K followers, [---] engagements


"vLLM-studio will fully support AMD GPUs and roc-m within the next [--] days. Development started today. Thanks to a generous grant from HotAisle. If youre interested in getting this running on your hardware DM. We have GPUs available for this. ssh https://t.co/R7PUEnvNDi We have GPUs available for this. ssh https://t.co/R7PUEnvNDi"  
[X Link](https://x.com/0xSero/status/2020237835912315081)  2026-02-07T20:46Z 10.6K followers, [----] engagements


"@dhh Youd need 25k in Apple hardware to cap at [--] tps or 100k in Nvidia for [--] tps"  
[X Link](https://x.com/0xSero/status/2020444851704070568)  2026-02-08T10:29Z 10.5K followers, [----] engagements


"@bernhard_me @AlexFinn You can rent a Ferrari while driving a Sedan"  
[X Link](https://x.com/0xSero/status/2020529059134869875)  2026-02-08T16:04Z 10.3K followers, [----] engagements


"GLM-4.7-Reap now on MLX Run GLM with tons of context on 128gb of memory (: https://huggingface.co/mlx-community/GLM-4.7-REAP-50-mxfp4/tree/main https://huggingface.co/mlx-community/GLM-4.7-REAP-50-mxfp4/tree/main"  
[X Link](https://x.com/0xSero/status/2007360740311617543)  2026-01-03T07:57Z 10.7K followers, 33.2K engagements


"Man what a model. I have not seen any mode below 200B act like this. It's really doing a good job in a pretty novel environment. The ZAI team is doing a great job I'm very happy they still open source all this. It's very impressive. https://huggingface.co/zai-org/GLM-4.7-Flash https://huggingface.co/zai-org/GLM-4.7-Flash"  
[X Link](https://x.com/0xSero/status/2013911569428586656)  2026-01-21T09:48Z 10.8K followers, 42.1K engagements


"I have another on the way a laptop which controls my 3d printer my AI rig and [--] more pewters in the closet. Not mentioning all the other pewters outside of my office. I need better cable management"  
[X Link](https://x.com/0xSero/status/2018628696479821927)  2026-02-03T10:12Z 10.7K followers, [---] engagements


"I am way more excited about GPT-5.3-Codex than any other model including the latest Opus. The speed of the OAI models has at least doubled in the last month and having the ability to go between 🤓 mode and 😎 mode is huge. I have crossed over to the right side where what I ask for is done without me having to spend 40% of every work week cleaning up messes. I still love Opus but for coding this lineup is the peak and pinnacle @synthwavedd https://t.co/cqVFd4yv58 @synthwavedd https://t.co/cqVFd4yv58"  
[X Link](https://x.com/0xSero/status/2019477220935246014)  2026-02-05T18:24Z 10.7K followers, [----] engagements


"The long awaited Clawdbot/Openclaw local model + home network guide. I showed many different usecases Ive been running since December. Ive chatted more with Openclaw than any other person in the last [--] days lol. https://youtu.be/IDi1HNym0u8si=NzTX1JQ714Tb1g1s https://youtu.be/IDi1HNym0u8si=NzTX1JQ714Tb1g1s"  
[X Link](https://x.com/0xSero/status/2020114769127059809)  2026-02-07T12:37Z 10.8K followers, 17.4K engagements


"New default model in openclaw is GTP-5.3-Codex in friendly personality mode. Since they doubled the rate limits its more than worth it. Model has gotten better at computer use"  
[X Link](https://x.com/0xSero/status/2020122231901241706)  2026-02-07T13:07Z 10.7K followers, [----] engagements


"Every [--] minutes I come across another kill application of AI. If thats what they meant by tired of winning Im raw by now Voxtral can now directly stream audio input into text output. Perfect for: - Live subtitles - Language learning apps - Note-taking tools - And more Made a demo for you to try directly on hugging face https://t.co/kHLqGo41Pq Voxtral can now directly stream audio input into text output. Perfect for: - Live subtitles - Language learning apps - Note-taking tools - And more Made a demo for you to try directly on hugging face https://t.co/kHLqGo41Pq"  
[X Link](https://x.com/0xSero/status/2020239139656872375)  2026-02-07T20:51Z 10.7K followers, [----] engagements


"- ROCm support - TTS / STT (call mode like chatgpt with ace-step & whisper-v3-large) - Image / Video support Also have something really special cooking you will be able to share a link with a friend/public broadcast for someone to share GPUs with to run larger model from vllm-studio. I can create a share link send it to a friend with gpus both AMD and Nvidia (soon Mac as well) and a large model will be split to pipeline parallelism with your friend. https://github.com/0xSero/vllm-studio/pull/54 https://github.com/0xSero/vllm-studio/pull/54"  
[X Link](https://x.com/0xSero/status/2021008385877897609)  2026-02-09T23:48Z 10.7K followers, [----] engagements


"Parchi my browser extension has a few new features. - Multi model orchestration - Use multiple models with various profiles to automate tasks let a non-vision model send screenshots to vision models - Deep research swarms - Relay mode + cli tool + skill - Allow any model to relay tasks or control the browser from the terminal - Useful for context sharding deep research pen-testing and QA - History + Improved settings and reading experience I have built this mostly around Kimi now as it's so good at this type of task. Step-3.5-flash also seems promising."  
[X Link](https://x.com/0xSero/status/2021010063083909303)  2026-02-09T23:55Z 10.8K followers, [----] engagements


"@ZenMagnets A bunch random tasks devops install and test x/y/z go update these configs. etc"  
[X Link](https://x.com/0xSero/status/2021293278168055984)  2026-02-10T18:40Z 10.7K followers, [----] engagements


"@sudoingX Whats the speed up from 1- [--] - [--] and are you using ddr4 or ddr5"  
[X Link](https://x.com/0xSero/status/2021502647803543866)  2026-02-11T08:32Z 10.7K followers, [---] engagements


"@itsjustmarky Run failed model architecture is weird will try again this weekend"  
[X Link](https://x.com/0xSero/status/2021508561550168296)  2026-02-11T08:56Z 10.8K followers, [--] engagements


"@VictorTaelin Swift is one of their weakest points actually. Its been miserable trying to build anything on IOS I dont care to learn it. React native would be better but then its react 😭 and you get useEffect slop everywhere. Its so over"  
[X Link](https://x.com/0xSero/status/2021509940054274052)  2026-02-11T09:01Z 10.7K followers, [---] engagements


"@cardsstacked Codex solves that btw. But yes I will do that tomorrow"  
[X Link](https://x.com/0xSero/status/2021692158932140426)  2026-02-11T21:05Z 10.7K followers, [--] engagements


"@royjossfolk I mean it works I'm just not prioritizing making it as seemless to setup as lmstudio. Try it out if u use claude code or codex or whatever even opencode with kimi for free tell them to set it up for you should work in [--] mins"  
[X Link](https://x.com/0xSero/status/2021711499488592156)  2026-02-11T22:22Z 10.7K followers, [---] engagements


"@royjossfolk it takes time once you get a venv that works you save it and lock it with a venv don't deal with it yourself run codex/opencode(free) and claude code to figure it out for you. The speed difference is insane 20k+ prefill speeds"  
[X Link](https://x.com/0xSero/status/2021720729059045804)  2026-02-11T22:59Z 10.7K followers, [--] engagements


"Let me make your life better. [--]. Open [--]. Sign in with X [--]. Click your PFP [--]. Click tasks [--]. Define a task and the interval to run it at. [--]. It'll run and email/notify you at your desired interval i use this for - Sales - Research - X content - etc. http://grok.com http://grok.com"  
[X Link](https://x.com/0xSero/status/2021630064610406773)  2026-02-11T16:59Z 10.8K followers, [----] engagements


"@luisgnet sir an IDE & Browser are the only acceptable GUIs"  
[X Link](https://x.com/0xSero/status/2021726770098307258)  2026-02-11T23:23Z 10.8K followers, [----] engagements


"Tomorrow's video will be a deep dive on: - GLM-5 - Kimi-K2.5 - MiniMax-M2.5 We're going deep"  
[X Link](https://x.com/0xSero/status/2021728786425712961)  2026-02-11T23:31Z 10.8K followers, 20K engagements


"I haven't used claude in forever"  
[X Link](https://x.com/0xSero/status/2021951029760405596)  2026-02-12T14:14Z 10.8K followers, [----] engagements


"Looking for sponsors to get GLM-5 running on ampere in 120gb for weights with 95% accuracy in agentic/coding workflows. sponsorship will buy compute for reproducing research I can't do this alone. https://x.com/0xSero/status/2021947003383718256s=20 It's working. But i am so compute starved right now I can only experiment with 7b models. [--] hours of work https://t.co/kxQttR8CTJ https://x.com/0xSero/status/2021947003383718256s=20 It's working. But i am so compute starved right now I can only experiment with 7b models. [--] hours of work https://t.co/kxQttR8CTJ"  
[X Link](https://x.com/0xSero/status/2021953188400685289)  2026-02-12T14:23Z 10.8K followers, [----] engagements


"@bridgemindai Speed is the issue right now. The model is excellent I had it do a complex migration from Hetzner & digital ocean onto AWS with maybe [--] prompts for [--] hours of work"  
[X Link](https://x.com/0xSero/status/2022001075033976888)  2026-02-12T17:33Z 10.8K followers, [----] engagements


"@Everlier FR though you can use even a trashy local model to do this"  
[X Link](https://x.com/0xSero/status/2022034117517619458)  2026-02-12T19:44Z 10.8K followers, [---] engagements


"If you want to use agents for building regardless of your experience level I highly recommend watching this video if you're more experienced skip to minute [--] https://youtu.be/VgR66ybAtdgsi=O_ovkZhMH7UKLmWW https://youtu.be/VgR66ybAtdgsi=O_ovkZhMH7UKLmWW"  
[X Link](https://x.com/0xSero/status/2022240882221875290)  2026-02-13T09:26Z 10.8K followers, [----] engagements


"Training Andrej Karpathys Nanochat on 4x RTX 3090s at 225W each: Step 2694/21400 (12.59% done) Loss: [----] Runtime: [----] hours Throughput: [----] tok/sec Temps: 52-57C VRAM: 19GB/24GB per card Total cost: 15$ at 55h Zero errors perfectly stable"  
[X Link](https://x.com/anyuser/status/1978664943827190116)  2025-10-16T03:30Z 10.8K followers, 149.9K engagements


"Opencode plugins are very powerful it makes me want to tweak it in a thousand ways. Check out: https://github.com/0xSero/open-queue https://github.com/0xSero/open-queue"  
[X Link](https://x.com/0xSero/status/2005395558039687343)  2025-12-28T21:48Z 10.8K followers, 29.6K engagements


"In pursuit of my goal to run Claude and GPT at home here's what I've learned about LLM compression. As my profile shows I use LLMs a lot. Probably in the top 0.1% of all individual users. Imagine being a dealer's #1 customer. If the dealer closes shop you're fucked. That's why I built this thing: --------------------------- I've tried every model locally every single one with over a thousand downloads on HuggingFace in the pursuit of the one. There are about [--] groupings at any point that are truly useful but they're all lacking when you try them for serious work at this scale. The most"  
[X Link](https://x.com/0xSero/status/2008122142613721242)  2026-01-05T10:23Z 10.8K followers, 34.2K engagements


"200$ Claude Max sub buys you 2500$ of api usage"  
[X Link](https://x.com/0xSero/status/2008156361440534833)  2026-01-05T12:39Z 10.8K followers, 294.7K engagements


"Kimi-K2.5 can use a computer. It ACES browser user look at that smooth navigation. Ok computer"  
[X Link](https://x.com/0xSero/status/2016266795355648489)  2026-01-27T21:47Z 10.8K followers, 48.4K engagements


"PSA buy yourself a T7 and backup your MacBook/Mac + move all your finances and gov data to another device"  
[X Link](https://x.com/0xSero/status/2017600786402795583)  2026-01-31T14:08Z 10.8K followers, 461.6K engagements


"Hey let me make your life better. Regardless whether you're a moltbook enjoyer or whether you use AI to do some agentic coding or you write TS/JS and use NPM. [--]. Go out today/tomorrow and buy some external SSD/NVMe [--]. Setup Time Machine and encrypt it [--]. Backup your computer [--]. If you have ANY crypto or keychain software get a hardware wallet/ubi key [--]. Use some old hardware for all financial records and family documents. Regardless of how risk averse you are we will see an increase in incredibly strange hacks security vulnerabilities and exploits. It doesn't matter how much you engage in AI"  
[X Link](https://x.com/0xSero/status/2017616359585149423)  2026-01-31T15:10Z 10.8K followers, 435.8K engagements


"🦞 I caved Openclaws is genuinely useful and impressive but it needs its own environment. 6-700$ is a good price for an always on assistant. I have a Linux box but I would rather get it all setup for the robot explicitly"  
[X Link](https://x.com/0xSero/status/2017699685239079423)  2026-01-31T20:41Z 10.8K followers, 40K engagements


"Do you want the best web-scraper out there [--]. Set Opus as main model [--]. Set haiku for subagents [--]. Turn on chrome plugin [--]. Give it a few search api envs costs cents [--]. Prompt it to batch scraping targets to haiku subagents [--]. Tell it to output JSON responses. Very good way to find data that's not easily accessible online they can try programmatic ways then use the browser if targets not found. https://twitter.com/i/web/status/2017907852120060389 https://twitter.com/i/web/status/2017907852120060389"  
[X Link](https://x.com/0xSero/status/2017907852120060389)  2026-02-01T10:28Z 10.8K followers, 82.8K engagements


"Let me save you hours of testing frontends. If you're ever working on a front-end instead of writing tests and adding puppeteer slop to your repo [--]. Get an llm to write you with whatever needs to be tested [--]. Copy that go to browser [--]. Open localhost with your selected app [--]. Use Claude Chrome Extension or Parchi [--]. Send it the prompt [--]. QA engineering there you go. Use models results and pass it back to your coding agent to fix whatever is flagged. http://expected.md http://expected.md"  
[X Link](https://x.com/0xSero/status/2018101731591176325)  2026-02-01T23:18Z 10.8K followers, 111.9K engagements


"Today's youtube video is about Zed IDE which I believe is the best for AI it's the only IDE that let's you manage multiple projects while you code still It's AI framework (ACP) is top tier gives you a cursor like exp for free I recommend it. https://youtu.be/1FUKYyEumrA https://youtu.be/1FUKYyEumrA"  
[X Link](https://x.com/0xSero/status/2020480312753422719)  2026-02-08T12:50Z 10.8K followers, 12.2K engagements


"I checked the OpenCode desktop app today and it has gotten significantly better. All the existing desktop apps don't support opening files and checking your file tree for some reason. Kudos to their team & Adam"  
[X Link](https://x.com/0xSero/status/2020480609148084654)  2026-02-08T12:51Z 10.8K followers, 14.9K engagements


"Guys I just tried Step-3.5-Flash locally. - 200B params - 11B active params - ON llama.cpp 600tps prefill 58tps generation - SOOOOO good. I had it setup my cliproxyapi and it did it [--] shot no support needed. Incredible model try it out"  
[X Link](https://x.com/0xSero/status/2020614920447176984)  2026-02-08T21:45Z 10.8K followers, 24.2K engagements


"Top open-weight model below 400B right now incredible. Very good at doing things doesn't make a mess. https://huggingface.co/stepfun-ai/Step-3.5-Flash https://huggingface.co/stepfun-ai/Step-3.5-Flash"  
[X Link](https://x.com/0xSero/status/2020620147803758803)  2026-02-08T22:05Z 10.8K followers, [----] engagements


"My only use for Claude now is Clawdbot it is literally just not the best at anything anymore just an overall good model. - Kimi-K2.5 is better at automating computer use especially browser use I am using Parchi more than Claude chrome extension & BETTER at frontend - GPT is an infinitely better programmer requires less handholding produces much better results - MiniMax-M2.* & GLM are cheaper and better than Sonnet I use both tenfold more than I use Sonnet I haven't touched Sonnet in months. - Grok is much better at finding information about people I use Grok chat more than Claude chat. Not"  
[X Link](https://x.com/0xSero/status/2021287509540421734)  2026-02-10T18:17Z 10.8K followers, 55.5K engagements


"@hlitoroy IDE / Terminal + any model == anything u want no UI needed https://steipete.me/posts/just-talk-to-it https://steipete.me/posts/just-talk-to-it"  
[X Link](https://x.com/0xSero/status/2021291419261083820)  2026-02-10T18:33Z 10.8K followers, [----] engagements


"Here's how I use agents to work on [--] projects a time I think this is a unique video no one else is doing it like this. I promise you'll double your outputs. https://youtu.be/VgR66ybAtdg https://youtu.be/VgR66ybAtdg"  
[X Link](https://x.com/0xSero/status/2021654377958875430)  2026-02-11T18:35Z 10.8K followers, [----] engagements


"My most delusional idea yet. Implement DSA support for Ampere sm86 architecture. Lets see how good these models really are at coding. Novel compression research + ampere support = Deepseek at home"  
[X Link](https://x.com/0xSero/status/2021865182776230072)  2026-02-12T08:33Z 10.8K followers, [----] engagements


"I gave my clanker a credit card hetzner CLI access namecheap an email a crypto wallet. Let's see how it goes. I used a digital card with 165$ monthly limit on it no damage it can do"  
[X Link](https://x.com/0xSero/status/2021881289322729787)  2026-02-12T09:37Z 10.8K followers, [----] engagements


"@MiniMax_AI Will you release the weights"  
[X Link](https://x.com/0xSero/status/2022001260287627470)  2026-02-12T17:34Z 10.8K followers, [----] engagements


"@PussyBlaster10k You dont always need megamind"  
[X Link](https://x.com/0xSero/status/2022054462446301262)  2026-02-12T21:05Z 10.8K followers, [--] engagements


"Best open weight labs: #1 MiniMax: they squeeze every last drop of intelligence out of such a small model #2 Kimi: Their latest model is so good at computer use I dont think most people have tried it in Parchi it doesnt fail. #3 GLM: This is probably the most intelligent but also the hardest to use similar to GPT Best closed weight labs: #1 OpenAI: - Spark is revolutionary - GPT codex is the best coder - GPT Pro is the best researcher Thats it folks we are out of LLMs my GPUs are itching for MiniMax open weights"  
[X Link](https://x.com/0xSero/status/2022102064902320607)  2026-02-13T00:14Z 10.8K followers, 14.4K engagements


"@johntheyoung Im on gpt pro Also Minimax"  
[X Link](https://x.com/0xSero/status/2022201937970766315)  2026-02-13T06:51Z 10.8K followers, [---] engagements


"@Rahatcodes Acktually MiniMax-M2.* runs at [--] tps on Mac Studios (:"  
[X Link](https://x.com/0xSero/status/2022206204639162759)  2026-02-13T07:08Z 10.8K followers, [---] engagements


"Let me make your life better. [--]. Go find an old laptop computer raspberry-pie laying around somewhere in your home. (Or your computer if you're brave) [--]. Install Clawdbot: [--]. Buy [--] or [--] of these subs. minimax is very fast and the most Claude like GLM is more GPT-like you can get a year of both for 140$ total [--]. Set it up with peekaboo give it a browser. [--]. Connect to TG or discord I use both I have been using this for over a month and it's been so incredibly useful I can work from anywhere and control my machine works great with local models too. - - https://z.ai/subscribe"  
[X Link](https://x.com/0xSero/status/2015094964498166055)  2026-01-24T16:10Z 10.8K followers, 322.3K engagements


"Hey let me make your life easier. [--]. Go to Tailscale site [--]. Install the desktop app & mobile app [--]. Hook them up together via vpn [--]. Go to Termius [--]. Install the mobile app [--]. Set up using your tailscale IP [--]. Now you can control your computer from phone w no exposed ports"  
[X Link](https://x.com/0xSero/status/2017674123565060375)  2026-01-31T18:59Z 10.8K followers, 113.1K engagements


"I am working on implementing a new quantization method that should reduce hardware needs by 92% with relatively low loss in performance Here's the research paper very interesting. If this works out we can run Deepseek on [--] rtx [----] https://arxiv.org/abs/2602.05269 https://arxiv.org/abs/2602.05269"  
[X Link](https://x.com/0xSero/status/2021250793345548417)  2026-02-10T15:51Z 10.8K followers, 58.4K engagements


"Did you buy a Mac Mini for openclaw Let me make your life better. [--]. Install Zed IDE [--]. settings.json and drop this: "use_system_window_tabs": true [--]. click + on the top add all your repos [--]. click + do add remote directory [--]. enter your ip or tailscale network ip [--]. use your mac from your macbook [--]. add any other computers control all of them from [--] machine without any issues in [--] window"  
[X Link](https://x.com/0xSero/status/2021710890974773307)  2026-02-11T22:20Z 10.8K followers, 49.6K engagements


"Throwing another Hail Mary out there to the world. I want to work on: [--]. Applying and scaling novel compression methods for LLMs [--]. Implementing kernels for DSA on ampere hardware This is my Ethereum address (works for all evm blockchains) Im happy to share a Solana/BTC and maybe go fund me or whatever: - 0xc0163E58648b247c143023CFB26C2BAA42C9d9A9 Funding this does not mean you will get anything but me trying to make this happen I have had some success you can view: Over half of my experiments fail miserably and I feel quite embarrassed because I am taking the build in public thing very"  
[X Link](https://x.com/0xSero/status/2021871381562950137)  2026-02-12T08:57Z 10.8K followers, [----] engagements


"Do you use AI for frontend development Let me make your life easier. Provide a simple scope of what you want to change and ask the model to generate [--] html files each with [--] variations of the feature. Cherry pick from samples. This saves countless hours reprompting"  
[X Link](https://x.com/0xSero/status/2021937607350116735)  2026-02-12T13:21Z 10.8K followers, 12K engagements


"This is what I have been waiting for codex go brrrr. If anyone says codex is too slow again 1000+ tokens a second going to bend space and time Hello GPT-5.3-Codex-Spark ✨ Our first real-time coding model. It is FAST. 1000+ tokens per second. Once you experience latency this low its hard to go back. This is an exciting first milestone in our partnership with @Cerebras. https://t.co/ieuDkJ1qLz Hello GPT-5.3-Codex-Spark ✨ Our first real-time coding model. It is FAST. 1000+ tokens per second. Once you experience latency this low its hard to go back. This is an exciting first milestone in our"  
[X Link](https://x.com/0xSero/status/2022043813041299940)  2026-02-12T20:23Z 10.8K followers, [----] engagements


"Dawg it's soooooooooo good"  
[X Link](https://x.com/0xSero/status/2022096040334438829)  2026-02-12T23:50Z 10.8K followers, 23.6K engagements


"@Rahatcodes Thats 25% faster than codex"  
[X Link](https://x.com/0xSero/status/2022206316924874955)  2026-02-13T07:08Z 10.8K followers, [--] engagements


"People approach a slow model (GPT [--] months ago) How they approach a fast model (say Sonnet) then get frustrated with it and dump it. The same is true vice versa. Fast/Small/Low-context/local-models can often do better at some types of tasks: [--]. QA / Testing [--]. Retrieval [--]. Real time work with a single file [--]. Cleaning up & debugging [--]. Prepping for a demo Etc. Slow models are good in that they let you walk away from the screen and still get things done often with higher accuracy. gpt-5.3 codex spark felt off to me which is why i didnt even post about it while the timeline was busy praising it"  
[X Link](https://x.com/0xSero/status/2022283864027107342)  2026-02-13T12:17Z 10.8K followers, [----] engagements

Limited data mode. Full metrics available with subscription: lunarcrush.com/pricing