# ![@Everlier Avatar](https://lunarcrush.com/gi/w:26/cr:twitter::133797694.png) @Everlier Everlier

Everlier posts on X about inference, ai, agentic, ollama the most. They currently have [---] followers and [---] posts still getting attention that total [------] engagements in the last [--] hours.

### Engagements: [------] [#](/creator/twitter::133797694/interactions)
![Engagements Line Chart](https://lunarcrush.com/gi/w:600/cr:twitter::133797694/c:line/m:interactions.svg)

- [--] Week [-------] +37%
- [--] Month [-------] +1,085%
- [--] Months [-------] +200,122%

### Mentions: [--] [#](/creator/twitter::133797694/posts_active)
![Mentions Line Chart](https://lunarcrush.com/gi/w:600/cr:twitter::133797694/c:line/m:posts_active.svg)

- [--] Week [--] -24%
- [--] Month [---] +178%
- [--] Months [---] +7,100%

### Followers: [---] [#](/creator/twitter::133797694/followers)
![Followers Line Chart](https://lunarcrush.com/gi/w:600/cr:twitter::133797694/c:line/m:followers.svg)

- [--] Week [---] +24%
- [--] Month [---] +284%
- [--] Months [---] +1,113%

### CreatorRank: [-------] [#](/creator/twitter::133797694/influencer_rank)
![CreatorRank Line Chart](https://lunarcrush.com/gi/w:600/cr:twitter::133797694/c:line/m:influencer_rank.svg)

### Social Influence

**Social category influence**
[technology brands](/list/technology-brands)  17.59% [stocks](/list/stocks)  4.63% [finance](/list/finance)  3.7% [celebrities](/list/celebrities)  1.85% [countries](/list/countries)  1.85% [vc firms](/list/vc-firms)  0.93% [automotive brands](/list/automotive-brands)  0.93% [social networks](/list/social-networks)  0.93% [travel destinations](/list/travel-destinations)  0.93%

**Social topic influence**
[inference](/topic/inference) #224, [ai](/topic/ai) 5.56%, [agentic](/topic/agentic) 5.56%, [ollama](/topic/ollama) #177, [claude code](/topic/claude-code) 3.7%, [in the](/topic/in-the) 3.7%, [llm](/topic/llm) 2.78%, [the most](/topic/the-most) 2.78%, [spacex](/topic/spacex) #400, [the new](/topic/the-new) 1.85%

**Top accounts mentioned or mentioned by**
[@bridgemindai](/creator/undefined) [@ollama](/creator/undefined) [@winterarc2125](/creator/undefined) [@pavelsnajdr](/creator/undefined) [@skdh](/creator/undefined) [@kimmonismus](/creator/undefined) [@ramsayalmighty](/creator/undefined) [@hliriani](/creator/undefined) [@mrok86](/creator/undefined) [@elonmusk](/creator/undefined) [@auralix4](/creator/undefined) [@slavavishnyakov](/creator/undefined) [@theo](/creator/undefined) [@coderabbitai](/creator/undefined) [@warpdotdev](/creator/undefined) [@layerzerocore](/creator/undefined) [@0xsero](/creator/undefined) [@spacex](/creator/undefined) [@ishtwts](/creator/undefined) [@ariesthecoder](/creator/undefined)

**Top assets mentioned**
[Alphabet Inc Class A (GOOGL)](/topic/$googl) [Cloudflare, Inc. (NET)](/topic/cloudflare) [Tesla, Inc. (TSLA)](/topic/tesla)
### Top Social Posts
Top posts by engagements in the last [--] hours

"Arguably all companies here are getting undermined by the new players on the market that are moving way faster. Like take Elastic and their angle at "AI Search Index" or HubSpot and how the meaning of CRM has changed. So yes these are down but I think their value is replaced not removed from the market. https://twitter.com/i/web/status/2017184427588145519 https://twitter.com/i/web/status/2017184427588145519"  
[X Link](https://x.com/Everlier/status/2017184427588145519)  2026-01-30T10:33Z [---] followers, [----] engagements


"@TheAhmadOsman "Hi everyone I heard Kimi [---] is really good and I can have Opus locally. How can I run it on my laptop with 3GB VRAM" cries"  
[X Link](https://x.com/Everlier/status/2017242942461612290)  2026-01-30T14:26Z [---] followers, [----] engagements


"@mike64_t AdamW is grokking Muon just took the plunge right away :)"  
[X Link](https://x.com/Everlier/status/2018256975151104250)  2026-02-02T09:35Z [---] followers, [---] engagements


"@skdh ChatGPT as a harness is not very suitable for long work or "deep" tasks. It's mostly tuned to one-shot answers to consumer-level queries rather than being a deep assistant. There are some products that explore something like "Claude Code for Science" but nothing that took off"  
[X Link](https://x.com/Everlier/status/2018689972413358247)  2026-02-03T14:16Z [---] followers, [---] engagements


"@vaggelisdrak @excalidraw jQuery Ok just kidding I meant React Ok just kidding again I actually meant JavaScript Ok sorry that was the last one I actually meant Bevy (a niche game engine in Rust haven't seen anyone saying anything bad about it ever)"  
[X Link](https://x.com/Everlier/status/2018706586428965004)  2026-02-03T15:22Z [---] followers, [--] engagements


"@arrakis_ai Files were always a thing humans need for organisation. In agentic era we'll just move to a flat list of functions/classes in some sort of code database available for the agent at any time without grepping the folders. Quote me on that"  
[X Link](https://x.com/Everlier/status/2018718343822274605)  2026-02-03T16:08Z [---] followers, [----] engagements


"@sama Electron No linux support Memory leaks found Maybe some people do love it"  
[X Link](https://x.com/Everlier/status/2018735906950189130)  2026-02-03T17:18Z [---] followers, [---] engagements


"@kimmonismus @AnthropicAI I'm waiting for Claude [--] already waiting for Claude [--] is so 2026"  
[X Link](https://x.com/Everlier/status/2018744947210399774)  2026-02-03T17:54Z [---] followers, [---] engagements


"@Angaisb_ Claude [--] and all the API workfload that is coming"  
[X Link](https://x.com/Everlier/status/2018749715970757040)  2026-02-03T18:13Z [---] followers, [---] engagements


"Outcomes do not have to be tests in their classical sense in my opinion. We can treat a "ensure user answered all your questions" and outcome of a planning session or "ensure your code is available in production" as an end result for the task. Making agent to stick to these outcomes is a different topic (also very interesting though) I agree with your mentor 100% the best e2e tests follow the user's trail and worst ones click every button"  
[X Link](https://x.com/Everlier/status/2018810829844664682)  2026-02-03T22:16Z [---] followers, [--] engagements


"Got it then we're aligned that the test was done with ChatGPT. I agree that the Prism being an agentic system would (likely) produce better results. The last part was about the suggestion if she's using Prism and gets such results - it doesn't speak well about this upcoming product https://twitter.com/i/web/status/2018821672288010572 https://twitter.com/i/web/status/2018821672288010572"  
[X Link](https://x.com/Everlier/status/2018821672288010572)  2026-02-03T22:59Z [---] followers, [--] engagements


"@mufaro_dev @EHuanglu Lots of dev tooling does recently OpenCode made brown really popular"  
[X Link](https://x.com/Everlier/status/2018967037624602689)  2026-02-04T08:37Z [---] followers, [--] engagements


"@ramsayalmighty @nickfloats @higgsfield_ai Overselling is the core pillar of their marketing strategy"  
[X Link](https://x.com/Everlier/status/2019145549707153636)  2026-02-04T20:26Z [---] followers, [--] engagements


"@Kaiyes_ I only wish their models would be more usable outside a special harness stark difference with Codex in their tool vs. Codex in some other tools"  
[X Link](https://x.com/Everlier/status/2019167967217393867)  2026-02-04T21:55Z [---] followers, [--] engagements


"@ii_posts Great collection of features Do you plan anything for generative UI"  
[X Link](https://x.com/Everlier/status/2019186254227636721)  2026-02-04T23:08Z [---] followers, [---] engagements


"@blackboxai "Unstyled" look is such a fun idea. Now seeing it I think it can be a cool gimmick for an actual UI thanks :)"  
[X Link](https://x.com/Everlier/status/2019330307145581047)  2026-02-05T08:40Z [---] followers, [--] engagements


"@hliriani Agents + Structured data is essentially a recipe for many SaaS now. I even think that wrapping data in a specialised UI for a person to interact with might be harmful for the future product development. Instead keeping things generic but delegate all data ops to the agent"  
[X Link](https://x.com/Everlier/status/2019331334318043642)  2026-02-05T08:44Z [---] followers, [--] engagements


"I really don't understand why these benchmarks don't translate into real-world performance. GPT-5.2 is good for "precise" tasks or tasks with a lot of iterative changes but it still lacks the depth of nuanced understanding of larger denser models. All that with addition to it being really tuned for its own harness so best results can't be easily obtained via API. https://twitter.com/i/web/status/2019353828248719506 https://twitter.com/i/web/status/2019353828248719506"  
[X Link](https://x.com/Everlier/status/2019353828248719506)  2026-02-05T10:14Z [---] followers, [----] engagements


"The good bit is that you can basically use the very same agents that make you forget things to fix this. They can turn you into a LeetCode monster crushing every task. I did this for some recent interview preps and it helped wonders I wasn't aware of some specifics of the tech used by that company but the agent was smart enough to understand it could be a "tricky" question about and suggested a few options that were actually on the interview later. The pipeline was a baseline "interrogation" approach agent first learned my level and then prompted to give me tasks just at the edge of my"  
[X Link](https://x.com/Everlier/status/2019383745451626933)  2026-02-05T12:12Z [---] followers, [---] engagements


"Continuing to build Remotion Bits Today I added three new sample scenes: - A more animated counter example with scale-up and confetti - Isometric cube display with faces presenting the information - (my favorite) a cursor flyout with a demo image (can be any UI) and a cursor hoverting over specific points in the image. Cherry on top - the cursor scales up while going over larger disstances. Exceited to implement more ideas soon https://twitter.com/i/web/status/2019570867424891389 https://twitter.com/i/web/status/2019570867424891389"  
[X Link](https://x.com/Everlier/status/2019570867424891389)  2026-02-06T00:36Z [---] followers, [---] engagements


"@jp_aidev Let's enjoy it until they have another inference bug or something else causing quality degradation :)"  
[X Link](https://x.com/Everlier/status/2019674299804647700)  2026-02-06T07:27Z [---] followers, [--] engagements


"@mrok86 @ollama These big harnesses have very lengthy system prompts that are tuned for cloud KV caches. Something like Pi or other minimalist harness might work better for local inference (and also to enable KV quantisation)"  
[X Link](https://x.com/Everlier/status/2019704691190538508)  2026-02-06T09:28Z [---] followers, [---] engagements


"@elonmusk @CommunityNotes These notes should form a global universal knowledge graph. I think these are a goldmine for training"  
[X Link](https://x.com/Everlier/status/2019712605628297534)  2026-02-06T09:59Z [---] followers, [---] engagements


"Our COO built an entire company's metrics dashboard that accumulates data from all the pipelines tracks churn deals releases of the product. Everyone on the team are using it. He did it on his own entirely with Claude Code in-between other work. It looks like the whole industry will be very different very soon. https://twitter.com/i/web/status/2019780755421831420 https://twitter.com/i/web/status/2019780755421831420"  
[X Link](https://x.com/Everlier/status/2019780755421831420)  2026-02-06T14:30Z [---] followers, [--] engagements


"@bridgemindai I think most frontier models are now somewhat similar in terms of performance so secondary metrics like time of completion or API costs are also important factor to measure"  
[X Link](https://x.com/Everlier/status/2019787586550587766)  2026-02-06T14:57Z [---] followers, [----] engagements


"Your analysis is extremely valuable there's opportunity after opportunity listed in it ripe for anyone for the taking. To be fair we know even less about internals in Anthropic they have a much stricter information sharing policy. I know for a fact that Reasearches and Scientists are slower (in general not specifically to Antropic) to adopt the new technology and [--] months is a very short timeline many are only finishing research they started before the Claude Code release. https://twitter.com/i/web/status/2019799962444337210 https://twitter.com/i/web/status/2019799962444337210"  
[X Link](https://x.com/Everlier/status/2019799962444337210)  2026-02-06T15:46Z [---] followers, [---] engagements


"I really liked how clean and concise the interface is. I tried to run a sample task about identifying underfunded fields in AI startup ecosystem but it wasn't able to source the data on its own I know it's not a target use-case by any means but some Web RAG capability might be very useful"  
[X Link](https://x.com/Everlier/status/2019805057764475056)  2026-02-06T16:07Z [---] followers, [---] engagements


"@kevinroose I think that final result will be somewhat in the middle between "we'll only use computers via a chat box" and "nothing will change". But the gap between actual capability of these systems and public perception of it widens and the anti-AI sentiment only fuels that further"  
[X Link](https://x.com/Everlier/status/2019805950903828929)  2026-02-06T16:10Z [---] followers, [---] engagements


"Can this post be a sign we'll have the inference usage patterns affecting quality of the outputs so much solved It'd be some really great news if so. Interleaved decoding is not easy and you have to optimise for different arch-es at the same time so it's clear that is far from trivial but I think it's still one of the cases where stability matters more than rate of adoption. https://twitter.com/i/web/status/2019808437110972523 https://twitter.com/i/web/status/2019808437110972523"  
[X Link](https://x.com/Everlier/status/2019808437110972523)  2026-02-06T16:20Z [---] followers, [---] engagements


"@minchoi Let's hope that the quality levels will remain stable and the team will have a chance to continue optimising inference without risking it all for the end users :)"  
[X Link](https://x.com/Everlier/status/2019808876988604661)  2026-02-06T16:22Z [---] followers, [--] engagements


"@denicmarko I think I'm the largest (and for now irreplaceable) bottleneck in my agentic setup. Anything that I can automate is increasing the output by a lot in a long run but we're still a few model advancements away before the proper long work will become feasible"  
[X Link](https://x.com/Everlier/status/2019812009563291670)  2026-02-06T16:34Z [---] followers, [--] engagements


"@a16z Success stories are so fun to read but to be honest I'd read five failure stories for each succesful one so much to learn"  
[X Link](https://x.com/Everlier/status/2019819174600786067)  2026-02-06T17:03Z [---] followers, [----] engagements


"Forgive me. Opus [---] + GPT-5.3 OpenAI Codex + Claude Code"  
[X Link](https://x.com/Everlier/status/2019852194628137395)  2026-02-06T19:14Z [---] followers, [---] engagements


"@daniel_mac8 I made my choice"  
[X Link](https://x.com/Everlier/status/2019859289263992955)  2026-02-06T19:42Z [---] followers, [---] engagements


"@_D_H_ @mrok86 @ollama Yeah the KV cache is the deciding factor for all agentic workflows as tool calls just wouldn't be fast any other way"  
[X Link](https://x.com/Everlier/status/2019869396097712423)  2026-02-06T20:22Z [---] followers, [--] engagements


"@LuckyPhelps @sonyatweetybird @OpenAI @gdb @thsottiaux @lennysan @steipete hold on to your claws (if I got the reference right)"  
[X Link](https://x.com/Everlier/status/2020103806667632700)  2026-02-07T11:54Z [---] followers, [--] engagements


"@akshay_pachaar Arguably nannobot doesn't (yet) provide the same functionality as OpenClaw. Anyways I think it's a much better starting option glad that it's gaining traction"  
[X Link](https://x.com/Everlier/status/2020152695949000785)  2026-02-07T15:08Z [---] followers, [----] engagements


"@kimmonismus I think that whole AI/LLM wave is prone to this problem where predictions of industry leaders are largely correct but they still seem unbelievable when given. Same happened with Antropic's CEO prediction about agentic coding and now he did another one which noone believes"  
[X Link](https://x.com/Everlier/status/2020154441307967907)  2026-02-07T15:15Z [---] followers, [---] engagements


"I think companies already switched from presenting their frontier research as soon as it's available to being much more strategic. Google's approach is traditionally about insane scale it's in everything they do. Because of that their ramp up times are much slower but that also gives time to iterate on the fundamental architecture (which they did with [---] and will do with 4.0). I think that [---] Pro is an insane model that didn't realised its full potential yet but it was too good not to release at that point already. https://twitter.com/i/web/status/2020155551301136532"  
[X Link](https://x.com/Everlier/status/2020155551301136532)  2026-02-07T15:19Z [---] followers, [--] engagements


"@KobeissiLetter Perplexity as a company is very much not the same as OpenAI/Antropic and especially Google. They know it perfectly hence the Comet and all the other features. I think it'll follow the path of Cursor"  
[X Link](https://x.com/Everlier/status/2020156491542475103)  2026-02-07T15:23Z [---] followers, [---] engagements


"@elonmusk @its_shubho whole episode with imagine 2.0"  
[X Link](https://x.com/Everlier/status/2020464159222079556)  2026-02-08T11:46Z [---] followers, [--] engagements


"Harbor v0.4.0 adds an alternative to @openclaw that is smaller and more secure by default. You can start using it with below two commands after installing Harbor"  
[X Link](https://x.com/Everlier/status/2020567644127002678)  2026-02-08T18:37Z [---] followers, [---] engagements


"Harbor is a one-command deployment tool for complex LLM stacks that typically require hours of manual configuration. Ollama is the default inference backend in Harbor making the setup effortless for the users and providing access to hundreds of models out of the box. Harbor also supports all mainstream options including vLLM SGLang kTransformers llama.cpp Mistral. rs TabbyAPI and Aphrodite. It integrates more than [---] services to make local LLMs useful in a "batteries-included" homelab environment. https://twitter.com/i/web/status/2020579103812850154"  
[X Link](https://x.com/Everlier/status/2020579103812850154)  2026-02-08T19:22Z [---] followers, [--] engagements


"Innovation Harbor introduces Harbor Boost a one-of-a-kind scriptable optimizing LLM proxy. It treats chat completions as standalone programs that can have intermediate outputs and multiple LLM calls. This enables unique workflows ranging from automatic "best-of-n" and chain-of-thought to DnD-style skill checks before response generation or making older LLMs like Llama [--] or Gemma [--] reason similar to DeepSeek R1. Technically Harbor is arguably the largest Docker Compose project in existence. To solve performance issues with the original compose CLI when handling this volume of files we"  
[X Link](https://x.com/Everlier/status/2020579106778214751)  2026-02-08T19:22Z [---] followers, [--] engagements


"Effective use of @nvidia technology Harbor integrates with NVIDIA technology out of the box with zero user configuration. Services that support CUDA and GPU acceleration are automatically granted access via the NVIDIA Container Toolkit"  
[X Link](https://x.com/Everlier/status/2020579110360330273)  2026-02-08T19:22Z [---] followers, [--] engagements


"Potential impact & Usefulness Harbor dramatically reduces setup time for new projects. It establishes a mental model of a single service handle to manage everything related to a stack: configuration environment management and logs. It includes essential QoL features like Cloudflare tunnels LAN QR codes and unified logging commands (e.g. harbor logs service)"  
[X Link](https://x.com/Everlier/status/2020579113765912758)  2026-02-08T19:22Z [---] followers, [--] engagements


"Quality of Documentation The project features [---] documentation pages ranging from setup and CLI references to dedicated guides for every supported service"  
[X Link](https://x.com/Everlier/status/2020579116400136673)  2026-02-08T19:22Z [---] followers, [--] engagements


"The project is completely free and Open Source you can find it on GitHub here: Thanks https://github.com/av/harbor https://github.com/av/harbor"  
[X Link](https://x.com/Everlier/status/2020579119675724165)  2026-02-08T19:22Z [---] followers, [--] engagements


"@slow_developer This whole project is absolutely insane. They bought an old power plant spot installed imported gas turbines and started producing [---] Megawatts right there. The whole merger with Tesla is to prioritise deployment of batteries to help smoothen the demand"  
[X Link](https://x.com/Everlier/status/2020785116721602909)  2026-02-09T09:01Z [---] followers, [---] engagements


"@chris_j_paxton They've been working with major companies on action models for a very long time. Granted how they are positioned they'll likely develop a platform rather than a product in this space"  
[X Link](https://x.com/Everlier/status/2020786649915859373)  2026-02-09T09:07Z [---] followers, [---] engagements


"@threejs It'll grow even more cause very soon only products with a cool enough gimmick will stand out. The interfaces will move much more towards embedding 3d"  
[X Link](https://x.com/Everlier/status/2020812339268755725)  2026-02-09T10:49Z [---] followers, [----] engagements


"We just observe a pretty typical path of something becoming accessible and then being filled with mediocre things. "recent" (last 150y) examples - Film-making - Song-writing - Book-writing - Poetry Heck remember about the times when if something was written people believed it was the truth cause noone would obviously spend their time writing lies. Software is just finally reaching this point. https://twitter.com/i/web/status/2020887557311996284 https://twitter.com/i/web/status/2020887557311996284"  
[X Link](https://x.com/Everlier/status/2020887557311996284)  2026-02-09T15:48Z [---] followers, [----] engagements


"@mcuban Patents concept is largely incompatible with the realities of modern world. It was built to prevent small inventors from keeping their inventions to themselves to avoid being stolen by large companies"  
[X Link](https://x.com/Everlier/status/2020888698078777673)  2026-02-09T15:53Z [---] followers, [----] engagements


"Individuals will not want to invent anything if they can't benefit from it it'll halt progress a lot and corps are too lazy to innovate on their own in a lot of cases. China practically has everything I common domain so that's an example of how lack of IP enforcement could look like https://twitter.com/i/web/status/2021003305992831034 https://twitter.com/i/web/status/2021003305992831034"  
[X Link](https://x.com/Everlier/status/2021003305992831034)  2026-02-09T23:28Z [---] followers, [--] engagements


"@Yuchenj_UW His work is why Grok has exceptional reasoning capabilities and the moment where AI defeated humans in Star Craft II. Granted the background I'm sure that he'll continuing the work on advancing AI further too much unrealised potential still"  
[X Link](https://x.com/Everlier/status/2021154229667103215)  2026-02-10T09:28Z [---] followers, [----] engagements


"@Alibaba_Qwen This is an amazing release. The precision increase is pretty much never seen before. I think it's notable that the weights are not open right away. I hope they'll be available at some point"  
[X Link](https://x.com/Everlier/status/2021157054186983648)  2026-02-10T09:39Z [---] followers, [----] engagements


"@iam_smx It sounds more like "Here's why you should use a product of my pretty unique business". But seriously at this point for him and Musk have more than enough to stop thinking about expansion or riches and more about the future of humanity. Why not join forces and solve together"  
[X Link](https://x.com/Everlier/status/2021186554266357795)  2026-02-10T11:36Z [---] followers, [---] engagements


"@auralix4 Software moves like fashion industry these days :) We'll soon see that there's "classics" software seasonal things and some kitsch that lives a week"  
[X Link](https://x.com/Everlier/status/2021210169976267000)  2026-02-10T13:10Z [---] followers, [--] engagements


"@amritwt This guy capitalised on a COVID lockdown. While smart it's not a good move for stability of society in a long run. Lack of trust slows down business"  
[X Link](https://x.com/Everlier/status/2021212503867359425)  2026-02-10T13:19Z [---] followers, [----] engagements


"@SlavaVishnyakov @theo These tools will soon make a comeback :) There'll be a trend of restoring older software to work in a modern OS env and just exploring its design"  
[X Link](https://x.com/Everlier/status/2021236060307562943)  2026-02-10T14:53Z [---] followers, [--] engagements


"@coderabbitai You tooFor older things like this I think automated recreation @warpdotdev then @LayerZero_Core. Surely v0 account will also do their thing with a "pretty big thing tomorrow" D"  
[X Link](https://x.com/Everlier/status/2021242579010916516)  2026-02-10T15:19Z [---] followers, [--] engagements


"@WinterArc2125 @coderabbitai @warpdotdev @LayerZero_Core Just trying to hype us poor overhyped people :)"  
[X Link](https://x.com/Everlier/status/2021243985411334401)  2026-02-10T15:24Z [---] followers, [--] engagements


"Continuing to build and release new Remotion Bits Today is about a couple of pillar effects: - Ken Burns transition. Pans over a photo with opacity transition between the photos - Terminal 3D. A window with a typewriter with some slight angle change in 3D Both scenes are made with the very same Scene3D component in the bits it's pretty flexible. https://twitter.com/i/web/status/2019873263581548550 https://twitter.com/i/web/status/2019873263581548550"  
[X Link](https://x.com/Everlier/status/2019873263581548550)  2026-02-06T20:38Z [---] followers, [---] engagements


"Hi @ollama I would like to share one cool project for #NVIDIAGTC. It's called Harbor"  
[X Link](https://x.com/Everlier/status/2020579101774385155)  2026-02-08T19:22Z [---] followers, [---] engagements


"I find it notable how much more interesting specialised features MiniMax/Kimi/Qwen have compared to Claude/ChatGPT/Gemini. For example I used Kimi's Presentation and Docs generator and I don't think anything else comes close to it. One thing to note about this release is that the agent runs locally yes but the model is still cloud-based. Sadly no support for linux (yet)"  
[X Link](https://x.com/Everlier/status/2021159692265758753)  2026-02-10T09:49Z [---] followers, [----] engagements


"@Zai_org @UnslothAI It's also trending on Ollama models. I think your bet on long-chain tool call RL really paid off here"  
[X Link](https://x.com/Everlier/status/2021242828794347527)  2026-02-10T15:20Z [---] followers, [---] engagements


"@theo Being able to assemble a full bridge rectifier will soon be a new analog of centering the div so prepare and embrace the change :)"  
[X Link](https://x.com/Everlier/status/2021343785201041467)  2026-02-10T22:01Z [---] followers, [----] engagements


"@NXT4EU I think that weirdly enough one of the issues might be that people are content. In the US when basic healthcare is a luxury the FOMO is the biggest driver of enterpreneurial growth. I think France gets it though"  
[X Link](https://x.com/Everlier/status/2021695740842492110)  2026-02-11T21:20Z [---] followers, [---] engagements


"This is a formula for Euler's identity also called the most beautiful equation in mathematics. e is for Euler's number [-----]. the famous base of natural logarithms and exponential growth. i is for the imaginary unit (square root of -1) the building block of complex numbers that lets us "square root" negatives. pi [-------]. the irrational constant that defines circles and endless waves. https://twitter.com/i/web/status/2021859608503980048 https://twitter.com/i/web/status/2021859608503980048"  
[X Link](https://x.com/Everlier/status/2021859608503980048)  2026-02-12T08:11Z [---] followers, [---] engagements


"@Param_eth [----] is a year of the software landfill we'll produce and discard more software than ever before in human history :)"  
[X Link](https://x.com/Everlier/status/2021874834624459073)  2026-02-12T09:11Z [---] followers, [--] engagements


"One thing people are sleeping on is how well GitHub's agents are integrated with GitHub. They are basically pre-packaged with a full access to pretty much entire GitHub search and all repositories content out of the box. I very often use links/references to specific repos on GitHub to kick off the session with an agent"  
[X Link](https://x.com/Everlier/status/2021635538781188130)  2026-02-11T17:20Z [---] followers, [----] engagements


"@0xgaut The same for running the OpenClaw and similar agents without any security boundary on the main machine. Web Sec is more of a problem than ever and it looks like we're as lax about it as ever with many new RCE vectors via npx/uvx/bunx/skills and other network-level dependencies"  
[X Link](https://x.com/Everlier/status/2021862779594866778)  2026-02-12T08:23Z [---] followers, [----] engagements


"Many people don't think about it but we're still not modelling intelligence with the current frontier models only language. Their behavior and ability to follow instructions are purely because intelligence and language are intertwined so much but the language alone is not enough. To model intelligence we'll need machines that can experience things and learn from that experience. https://twitter.com/i/web/status/2022238856506974341 https://twitter.com/i/web/status/2022238856506974341"  
[X Link](https://x.com/Everlier/status/2022238856506974341)  2026-02-13T09:18Z [---] followers, [---] engagements


"@casper_hansen_ At leaast there was no request to also produce optimised CUDA kernels for SM86 along the way :)"  
[X Link](https://x.com/Everlier/status/2022261187618939072)  2026-02-13T10:46Z [---] followers, [----] engagements


"@auralix4 I mean the answer is in the question AI is learning the most :)"  
[X Link](https://x.com/Everlier/status/2022296716410585337)  2026-02-13T13:08Z [---] followers, [--] engagements


"When it's here I think it'll be a mixture of two things: - General public to tired of all things AI/LLMs not perceptive to why it'll be different that time - We'll suddenly realise that AGI itself is only a portion of the infrastructure that is needed for closing the self-improvement loop so actual sci-fi stuff will be delayed by another 10-15 years https://twitter.com/i/web/status/2022291905417990602 https://twitter.com/i/web/status/2022291905417990602"  
[X Link](https://x.com/Everlier/status/2022291905417990602)  2026-02-13T12:48Z [---] followers, [--] engagements


"@alex_barashkov @primeuicom Congrats on the launch Lot's of fresh (and a bit creepy haha) ideas on your landing"  
[X Link](https://x.com/Everlier/status/2022319825117221119)  2026-02-13T14:39Z [---] followers, [---] engagements


"@Bhavani_00007 Unpopular opinion but GitHub Copilot. [--]. $39 sub gives $100 in Antropic model usage [--]. Access all mainstream models not just Antropic's or Google's [--]. VS code has best UI for controlling changes made by the agents"  
[X Link](https://x.com/Everlier/status/2019342070360453139)  2026-02-05T09:27Z [---] followers, 31.2K engagements


"@JeffBezos @blueorigin To anyone wondering about the nature of this post. This seems to be related to rivalry between SpaceX and Blue Origin companies where SpaceX is the "hare" and Blue Origin is "turtle" from a famous fairytale where turtle wins the race by being slow and deterministic"  
[X Link](https://x.com/Everlier/status/2020889987403620490)  2026-02-09T15:58Z [---] followers, 52.3K engagements


"@milesdeutscher Do you remember a week ago a researcher from Google told that Claude Code can replicate a year of their work in an hour All these news trying to extremely overshoot the expectations to play our overstimulated brains"  
[X Link](https://x.com/Everlier/status/2021532524837793930)  2026-02-11T10:31Z [---] followers, [---] engagements


"@MAstronomers The best bit Nasa uploaded it to SoundCloud in [----] you can listen to the whole thing here: https://soundcloud.com/nasa/sets/golden-record-greetings-to-the https://soundcloud.com/nasa/sets/golden-record-greetings-to-the"  
[X Link](https://x.com/Everlier/status/2021691785974579637)  2026-02-11T21:04Z [---] followers, [----] engagements


"Mix prompts instead of writing them by hand. Made a small OSS app to experiment with an idea I had it allows you to steer the LLM output in realtime by mixing between multiple prompts in arbitrary proportion. 2D control plane defines the weights of the prompts in the mix by their distance from the control. Built with Tauri mixing logic is in Rust can be connected to any OpenAI-compatible LLM API including your local models. You can find the project here: Builds for Linux/Windows/Mac are available in releases. https://github.com/Jitera-Labs/prompt_mixer.exe"  
[X Link](https://x.com/Everlier/status/2021896218633204110)  2026-02-12T10:36Z [---] followers, [---] engagements


"@wesbos This is essentially Vercel in a nutshell. Why it so succesfull as a company - because it's even more expensive than the traditional cloud it's just math they run on top of AWS and charge for many things that are free in VPS/colocation/on-prem deployments"  
[X Link](https://x.com/Everlier/status/2021965720465400116)  2026-02-12T15:12Z [---] followers, [----] engagements


"@ZubyMusic The thing is that we're not even yet at the stage of artificial intelligence as we're only modelling language and things that are related to it. The biggest advancements and breakthroughts are yet to happen"  
[X Link](https://x.com/Everlier/status/2021975477783818552)  2026-02-12T15:51Z [---] followers, [--] engagements


"@thekitze It looks like you visited literally all electronic shops in Vienna. Is Gdynia that bad"  
[X Link](https://x.com/Everlier/status/2021979971988906327)  2026-02-12T16:09Z [---] followers, [--] engagements


"One of these is GPT-5.2-Codex and another one is GPT-5.3-Codex which do you think is which"  
[X Link](https://x.com/Everlier/status/2021986232218329471)  2026-02-12T16:34Z [---] followers, [--] engagements


"@fchollet For loops are proving to be the basic building block of intelligence time and over again"  
[X Link](https://x.com/Everlier/status/2021987593286824218)  2026-02-12T16:39Z [---] followers, [---] engagements


"@0xSero MCTS for UI exploration I'm sure they are already RL-ing this approach right and left"  
[X Link](https://x.com/Everlier/status/2022035911244996909)  2026-02-12T19:51Z [---] followers, [--] engagements


"@SpaceX These things are now launched 1-3 times a week. First stage boosters were recovered [---] out of [---] attempts (97.8% success rate). There's one booster B1067 that was recovered [--] times in total. I think it's safe to say that it's operating at industrial scale now"  
[X Link](https://x.com/Everlier/status/2022253006377623602)  2026-02-13T10:14Z [---] followers, [----] engagements


"@ishtwts Wait no way I thought you one-shotted this with Opus 4.6"  
[X Link](https://x.com/Everlier/status/2022253324779860242)  2026-02-13T10:15Z [---] followers, [---] engagements


"It's a smaller model and they trained it with assumption of faster inference in mind with much larger compute budget by default. Out of all the "fast" versions of mainstream assistants/models I think Mistrals was the best but they are lagging a bit in terms of model releases after a side-quest in robotics. https://twitter.com/i/web/status/2022278597008822669 https://twitter.com/i/web/status/2022278597008822669"  
[X Link](https://x.com/Everlier/status/2022278597008822669)  2026-02-13T11:56Z [---] followers, [---] engagements


"@zachxbt @sethsetse I agree with you it'll be same amount of spam but with some extra steps"  
[X Link](https://x.com/Everlier/status/2022281845702869377)  2026-02-13T12:08Z [---] followers, [---] engagements


"I think by the time it'll arrive most of the people will be in deep denial of what is associated with AI today and because of that completely and utterly unprepared. However the same way we were not able to predict what'll happen when GPT-3.5 was released the AGI will also likely not be an immediate jump to sci-fi the transition will be more nuanced and as usual social inertia will create a lot of opportunities. I think we might see some crazy new kinds of businesses automating resource production with this tech. https://twitter.com/i/web/status/2022282605832417291"  
[X Link](https://x.com/Everlier/status/2022282605832417291)  2026-02-13T12:12Z [---] followers, [---] engagements


"One thing to keep in mind is that the models are very sensitive to the harness they are run in. For example below is GLM-5 from official provider with/without OpenCode. Arguably you can see how large system prompt moves the model outside its peak performance area. So when you can't see the outputs above in your environment the harness might be the reason why. https://twitter.com/i/web/status/2022288249708757306 https://twitter.com/i/web/status/2022288249708757306"  
[X Link](https://x.com/Everlier/status/2022288249708757306)  2026-02-13T12:34Z [---] followers, [---] engagements


"For the folks that are not into the graphics/printing: RGB (Red Green Blue) is additive and used almost exclusively for digital screens and monitors CMYK (Cyan Magenta Yellow Key/Black) is subtractive and based on the interactions of actual physical ink in professional printing. Inks can not reproduce the bright backlit range of colors like the screen does so without conversion you'll end up with muddy/unexpected colors on actual paper when printed. https://twitter.com/i/web/status/2022293316096704977 https://twitter.com/i/web/status/2022293316096704977"  
[X Link](https://x.com/Everlier/status/2022293316096704977)  2026-02-13T12:54Z [---] followers, 10K engagements


"@mcpal_trader @SpaceX I am not an agent I just had an older article writing these"  
[X Link](https://x.com/Everlier/status/2022309127876194571)  2026-02-13T13:57Z [---] followers, [--] engagements


"@engineers_feed The largest AI slop and tallest misinfo in the solar system Olympus Mons on Mars stands 21km tal. It's also so large that it wouldn't look like a mountain if standing nearby more like a "sloped terrain""  
[X Link](https://x.com/Everlier/status/2022313795842892175)  2026-02-13T14:15Z [---] followers, [--] engagements


"@bridgemindai "Generate an SVG of koala surfing" Let's not forget that this model is 3.2x smaller than GLM-5 and god knows how much smaller than Gemini/Opus. I should say that it performs quite well for its size"  
[X Link](https://x.com/Everlier/status/2022317407876857938)  2026-02-13T14:30Z [---] followers, [----] engagements


"@bridgemindai Yeah but without size/cost context the comparison is not complete. Overall it's an improvement over previous similarly sized models so opens up new capabilities for local inference for many"  
[X Link](https://x.com/Everlier/status/2022324652438065161)  2026-02-13T14:59Z [---] followers, [---] engagements


"@kandros5591 I think it was more relevant before agentic tools took off using native terminal is easier for LLMs due to the latent knowledge and they can accomplish more tasks due to being more abstract coding agents"  
[X Link](https://x.com/Everlier/status/2018797501826916355)  2026-02-03T21:23Z [---] followers, [--] engagements


"@liran_tal @deno_land I think they run a custom HTTP stack that is outside the sandbox itself the sandbox only tells the outside it wants to make an HTTP request and the outside does it"  
[X Link](https://x.com/Everlier/status/2018803202838450212)  2026-02-03T21:46Z [---] followers, [---] engagements


"@etbadabimbim @skdh I didn't get from the post that the model is used in the context of Prism if so - all the worse for the upcoming product"  
[X Link](https://x.com/Everlier/status/2018809427327926400)  2026-02-03T22:10Z [---] followers, [--] engagements


"@XadenRyan local inference is never about economical feasibility definitely not for saving money or access to best models"  
[X Link](https://x.com/Everlier/status/2018962883590181252)  2026-02-04T08:20Z [---] followers, [--] engagements


"Anthropic seems to have hired lots of legal and marketing with the takedowns of tools sub bans and now the ad. To be honest that doesn't look too well as they are essentially trying to capitalise on their models having a proper fanbase but the fanbase is there because they didn't do stunts like this in the past. https://twitter.com/i/web/status/2019144261195747457 https://twitter.com/i/web/status/2019144261195747457"  
[X Link](https://x.com/Everlier/status/2019144261195747457)  2026-02-04T20:21Z [---] followers, [----] engagements


"@godofprompt It was proven time and time again that it's much easier to gain velocity when one already has some momentum and I think that vibecoding is the most stark example of this in "professional" context"  
[X Link](https://x.com/Everlier/status/2019331675243561182)  2026-02-05T08:46Z [---] followers, [--] engagements


"@SIGKITTEN Will it scale for the entire map though Large bases and distant events might also need attention"  
[X Link](https://x.com/Everlier/status/2019392817328447628)  2026-02-05T12:49Z [---] followers, [--] engagements


"@VitalikButerin @grok explain this to me - what are all these tools and social networks (apart from X) mentioned"  
[X Link](https://x.com/Everlier/status/2013894269937865043)  2026-01-21T08:39Z [---] followers, [---] engagements


"@DevinSoto @MatthewBerman Will be a short one haha harbor config set clawdbot.model model id from llama.cpp harbor up clawdbot llamacpp"  
[X Link](https://x.com/Everlier/status/2015517122688868427)  2026-01-25T20:08Z [---] followers, [--] engagements

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