# ![@jackhnels Avatar](https://lunarcrush.com/gi/w:26/cr:twitter::2011536782496600064.png) @jackhnels Jack Nelson 🦞

Jack Nelson 🦞 posts on X about openclaw, ai, if you, model the most. They currently have [-------] followers and [--] posts still getting attention that total [-------] engagements in the last [--] hours.

### Engagements: [-------] [#](/creator/twitter::2011536782496600064/interactions)
![Engagements Line Chart](https://lunarcrush.com/gi/w:600/cr:twitter::2011536782496600064/c:line/m:interactions.svg)

- [--] Week [------] +26,512%

### Mentions: [--] [#](/creator/twitter::2011536782496600064/posts_active)
![Mentions Line Chart](https://lunarcrush.com/gi/w:600/cr:twitter::2011536782496600064/c:line/m:posts_active.svg)


### Followers: [-------] [#](/creator/twitter::2011536782496600064/followers)
![Followers Line Chart](https://lunarcrush.com/gi/w:600/cr:twitter::2011536782496600064/c:line/m:followers.svg)

- [--] Week [--] +253%

### CreatorRank: [-------] [#](/creator/twitter::2011536782496600064/influencer_rank)
![CreatorRank Line Chart](https://lunarcrush.com/gi/w:600/cr:twitter::2011536782496600064/c:line/m:influencer_rank.svg)

### Social Influence

**Social category influence**
[technology brands](/list/technology-brands)  [finance](/list/finance)  [stocks](/list/stocks)  [exchanges](/list/exchanges)  [financial services](/list/financial-services) 

**Social topic influence**
[openclaw](/topic/openclaw) #575, [ai](/topic/ai), [if you](/topic/if-you), [model](/topic/model), [anthropic](/topic/anthropic) #1559, [in the](/topic/in-the), [code](/topic/code), [claude code](/topic/claude-code) #724, [token](/topic/token), [codex](/topic/codex)

**Top accounts mentioned or mentioned by**
[@openclaw](/creator/undefined) [@grok](/creator/undefined) [@nummanali](/creator/undefined) [@kimimoonshot](/creator/undefined) [@minimaxai](/creator/undefined) [@zaiorg](/creator/undefined) [@openrouterai](/creator/undefined) [@vercel](/creator/undefined) [@ollama](/creator/undefined) [@nanogptcom](/creator/undefined) [@gatewayzai](/creator/undefined) [@iyoushetwt](/creator/undefined) [@legendaryy](/creator/undefined) [@chatgpt21](/creator/undefined) [@noahepstein](/creator/undefined) [@michaelclond](/creator/undefined) [@kobeissiletter](/creator/undefined) [@abhagsain](/creator/undefined) [@aarondfrancis](/creator/undefined) [@calebwinston](/creator/undefined)
### Top Social Posts
Top posts by engagements in the last [--] hours

"HOLY SHIT Anthropic Just Triggered a $285B Market Crash 😳 Bloomberg just reported Anthropic released a new AI tool that caused: $285 billion erased from software finance and asset management stocks 6% plunge in Goldmans software basket (worst since April) 7% collapse in the financial services index Nasdaq tanked 2.4% at its lowest This is MASSIVE. The market straight up panicked over an AI automation tool. If youre in software legal or IT services this changes everything. You just havent felt it yet. On Jan [--] Anthropic quietly shipped [--] plugins for Claude Cowork. Not a new model. Just"  
[X Link](https://x.com/jackhnels/status/2019182978476183889)  2026-02-04T22:55Z [--] followers, [--] engagements


"@DeFi_Hanzo I built these repos to help build better openclaw bots: Listening improvements: Best memory possible: Get him to work 24/7: https://github.com/Jacknelson6/nightcrew https://github.com/Jacknelson6/owl-brain https://github.com/Jacknelson6/clawdbot-atlas-kit https://github.com/Jacknelson6/nightcrew https://github.com/Jacknelson6/owl-brain https://github.com/Jacknelson6/clawdbot-atlas-kit"  
[X Link](https://x.com/jackhnels/status/2019202975768670636)  2026-02-05T00:14Z [--] followers, [---] engagements


"@KobeissiLetter @grok what is sanae takaichis stance on japans inflation and the US treasury bond market"  
[X Link](https://x.com/jackhnels/status/2019471937517371779)  2026-02-05T18:03Z [--] followers, [--] engagements


"@KobeissiLetter @grok where do things go next"  
[X Link](https://x.com/jackhnels/status/2019740903208845794)  2026-02-06T11:52Z [--] followers, [--] engagements


"@Mr_Derivatives @grok if this is true how do I capitalize"  
[X Link](https://x.com/jackhnels/status/2019745389130195050)  2026-02-06T12:10Z [--] followers, [--] engagements


"I completely agree here. People are focusing too much on what's the best tech and just not building. People need to just get their feet wet and get to work. Comparing Opus [---] and Minimax M2.5 and Kimi K2.5 and GPT [---] Spark are fun but it does absolutely nothing. You would get so much more done if you just started with the worst one. https://twitter.com/i/web/status/2022833752246796306 https://twitter.com/i/web/status/2022833752246796306"  
[X Link](https://x.com/jackhnels/status/2022833752246796306)  2026-02-15T00:42Z [--] followers, [---] engagements


"Deep think is fantastic Ive used it quite a bit for deep dives on specific topics. That being said it is atrociously slow for a very similar output to opus [---]. I know the benchmarks are in favor of gemini but in the real world speed will always be more important than small improvements. https://twitter.com/i/web/status/2022251768709587381 https://twitter.com/i/web/status/2022251768709587381"  
[X Link](https://x.com/jackhnels/status/2022251768709587381)  2026-02-13T10:09Z [--] followers, [---] engagements


"@iyoushetwt After Brave's API Upgrade yesterday it pretty much eliminated it entirely"  
[X Link](https://x.com/jackhnels/status/2022407742540124526)  2026-02-13T20:29Z [--] followers, 24.7K engagements


"@0xSero Hey on a head-to-head comparison I prefer Claude code over GPT Codex Codex may be fast but it's not as intelligent Plus having Opus [---] whenever I just wanna chat is perfect for Openclaw"  
[X Link](https://x.com/jackhnels/status/2022835923470229773)  2026-02-15T00:50Z [--] followers, [--] engagements


"@abhagsain @aarondfrancis Look into threading. You can thread conversations with Claude Code and Codex together and have them work together before you even start. They can essentially plan before you even begin"  
[X Link](https://x.com/jackhnels/status/2022836232124841991)  2026-02-15T00:51Z [--] followers, [---] engagements


"@maximelabonne Looks like GPT-3 flash preview might be the best between cost and performance Pretty sure it's like 10x cheaper than Opus 4.6"  
[X Link](https://x.com/jackhnels/status/2022839491338088937)  2026-02-15T01:04Z [--] followers, [---] engagements


"When you drive your car do you make your own gas My point is that as companies like OpenAI Anthropic Minimax and others scale even more cost will go down over time even if that's not the case initially. Capitalism rewards those who make things more accessible. There may be a period of time where local models can be helpful but nothing will outpace hundreds of thousands of NVIDIA processors. Not even your Mac Mini. https://twitter.com/i/web/status/2022743947567681990 https://twitter.com/i/web/status/2022743947567681990"  
[X Link](https://x.com/jackhnels/status/2022743947567681990)  2026-02-14T18:45Z [--] followers, [---] engagements


"@thesayannayak @claudeai @Kimi_Moonshot Just start they'll both get the job done. technically Opus [---] is better but that doesn't matter. Kimi K2.5 is cheaper and it has its place. If you can't afford Opus [---] then use Kimi K2.5. Just start building"  
[X Link](https://x.com/jackhnels/status/2022833922308972966)  2026-02-15T00:42Z [--] followers, [--] engagements


"@openclaw Valentines date with significant other new openclaw update testing"  
[X Link](https://x.com/jackhnels/status/2022881853770785088)  2026-02-15T03:53Z [--] followers, [----] engagements


"Openclaw BURNS through your claude tokens like crazy. Heres how Im saving thousands on token costs now: Codex CLI - $1.50/M input $6/M output. Runs locally uses your existing ChatGPT subscription and handles coding tasks that would eat through opus usage. Kimi K2.5 - $0.60/M input $3/M output. [--] trillion parameters but only activates 32B per query (MoE architecture). 100% open source too. Benchmarks competitive with models 10x the price. For context: Claude Opus is $15/$75 per million tokens. You don't need the most expensive model for every task. Route simple stuff to cheaper models save the"  
[X Link](https://x.com/anyuser/status/2021897441323168176)  2026-02-12T10:41Z [--] followers, [---] engagements


"If Opus [---] is the only model you're using for your Openclaw what are you even doing With Opus [---] you have two options: [--]. Get limited even with a $200/month plan. [--]. Spend thousands a month on Anthropic API This is the part where I urge you to explore other models especially ones that can be run on a 3rd party API provider https://twitter.com/i/web/status/2021995483216920706 https://twitter.com/i/web/status/2021995483216920706"  
[X Link](https://x.com/anyuser/status/2021995483216920706)  2026-02-12T17:11Z [--] followers, [---] engagements


"@Legendaryy @grok SWE benchmarks for Opus [---] vs glm [--] vs kimi k2.5"  
[X Link](https://x.com/jackhnels/status/2021999422503321956)  2026-02-12T17:26Z [--] followers, [---] engagements


"@grok @Legendaryy @grok how does agent swarms factor"  
[X Link](https://x.com/jackhnels/status/2021999899173433453)  2026-02-12T17:28Z [--] followers, [--] engagements


"@grok @Legendaryy @grok how does the sota mini max compare Does it also have agent swarms"  
[X Link](https://x.com/jackhnels/status/2022000468936077537)  2026-02-12T17:30Z [--] followers, [--] engagements


"@chatgpt21 Woah. do we know how the speed compares to opus [---] Or claude code"  
[X Link](https://x.com/jackhnels/status/2022019471490396517)  2026-02-12T18:46Z [--] followers, [----] engagements


"@chatgpt21 Yeah I just saw that it was like [----] tokens a second insanity I can't wait till this comes to plus. I'm trying out Codex right now and I feel like it majorly lacks in speed. Another big thing is that I've seen the accuracy really start to lose when it comes to spark"  
[X Link](https://x.com/jackhnels/status/2022071558391713942)  2026-02-12T22:13Z [--] followers, [---] engagements


"My advice as a founder and head of creative for social media generating billions of views for our clients: - Start now and don't stop. - It's only getting harder to get into organic social. - The more experience you gain the better your content gets. - The faster you can grow. - This skill isn't going anywhere and it's going to be necessary very soon. https://twitter.com/i/web/status/2022072690798919929 https://twitter.com/i/web/status/2022072690798919929"  
[X Link](https://x.com/jackhnels/status/2022072690798919929)  2026-02-12T22:17Z [--] followers, [--] engagements


"@virattt Yeah I've been going head-to-head with Codex and Opus. I think Opus very much excels in the design department and overall architecture but Codex does a better job at reasoning and thinking through things clearly and avoiding some of the pitfalls that Opus Falls into"  
[X Link](https://x.com/jackhnels/status/2022073003321045486)  2026-02-12T22:19Z [--] followers, [---] engagements


"@markbuildsbrand What part of Opus [---] do you like so much better than Opus [---] I feel like the difference there is pretty marginal"  
[X Link](https://x.com/jackhnels/status/2022073123185864798)  2026-02-12T22:19Z [--] followers, [---] engagements


"@CalebWinston This is epic. Please send over Excited to see how it works. Is that $0.15 coming from x402 Or model API cost Keep it up Recon looks awesome. Glad to see it as a CLI too"  
[X Link](https://x.com/jackhnels/status/2022105572863283306)  2026-02-13T00:28Z [--] followers, [--] engagements


"@atomicbot_ai Charging via x402 If not that might be the move"  
[X Link](https://x.com/jackhnels/status/2022106121826447469)  2026-02-13T00:30Z [--] followers, [---] engagements


"They're pairing super nicely together. Look into X4O2. Crypto is one of the most natural payment forms to pair with AI. This is really saying something for the future of the world since the two most disruptive technologies that we've encountered over the past two decades are built almost to pair with each other. https://twitter.com/i/web/status/2022137665861116049 https://twitter.com/i/web/status/2022137665861116049"  
[X Link](https://x.com/jackhnels/status/2022137665861116049)  2026-02-13T02:36Z [--] followers, [--] engagements


"@bridgemindai Where does GPT [---] come into play I have a theory that GPT [---] might beat Claude Opus [---]. For Openclaw this ranking makes it pretty clear that Mini Max M [---] is now the state of the art when it comes to models"  
[X Link](https://x.com/jackhnels/status/2022138878510108764)  2026-02-13T02:40Z [--] followers, [---] engagements


"@askOkara Love this comparison. When it comes to real world application speed will always be more important"  
[X Link](https://x.com/jackhnels/status/2022251105582006567)  2026-02-13T10:06Z [--] followers, [---] engagements


"RLM is very similar to using sub-agents but it differs from the overall purpose. An RLM uses Python to look for keywords in a large context. The orchestrator or the RLM sends out many LLMs to parse specific parts of that large and those many LLMs report back with a score to the RLM. The orchestrator LLM looks at that score and if it feels like it did not get enough info it will send that LLM back out to get more info. The reason this works more effectively and differently than a typical sub-agent is that sub-agents do not have that REPL filter. It is simply LLM-to-sub-agent versus"  
[X Link](https://x.com/jackhnels/status/2022276623874035925)  2026-02-13T11:48Z [--] followers, [--] engagements


"@feifei_qiu Moats. Build user-base and have them invest in your software. Make it harder for them to leave than to stay and pay a premium"  
[X Link](https://x.com/jackhnels/status/2022386084089598441)  2026-02-13T19:03Z [--] followers, [---] engagements


"@ZenMagnets @godofprompt Totally but that's where you can use something else to do the multi modal interpreting"  
[X Link](https://x.com/jackhnels/status/2022406524694598012)  2026-02-13T20:24Z [--] followers, [--] engagements


"@Mateovogt @iyoushetwt It honestly is. Opus [---] still better but its 20x cheaper sooo"  
[X Link](https://x.com/jackhnels/status/2022432313918001407)  2026-02-13T22:06Z [--] followers, [----] engagements


"@NoahEpstein_ @ManusAI Couple different things: automation for operations inside my marketing agency as well as some coding"  
[X Link](https://x.com/jackhnels/status/2022437885568315516)  2026-02-13T22:29Z [--] followers, [---] engagements


"@HappyGezim @cristthian @openclaw Totally agree although its pointless to run opus or any expensive models through API. Much better to use oauth tokens instead"  
[X Link](https://x.com/jackhnels/status/2022521557533298729)  2026-02-14T04:01Z [--] followers, [--] engagements


"@EntrepGrow Builders teach other builders and hire them for a fraction of the cost while everyone else is stuck hiring devs who already learned claude code and codex. As a builder you have knowledge the rest of the population doesnt have. Use it"  
[X Link](https://x.com/jackhnels/status/2022523698641592583)  2026-02-14T04:10Z [--] followers, [--] engagements


"@BHolmesDev Eh codex is slow unintelligent sometimes and chooses the wrong path often. I choose cc"  
[X Link](https://x.com/jackhnels/status/2022523968952111215)  2026-02-14T04:11Z [--] followers, [--] engagements


"@0xDesigner This is exactly why more people need to learn about recursive language models and the beauty of threading"  
[X Link](https://x.com/jackhnels/status/2022534872900739176)  2026-02-14T04:54Z [--] followers, [--] engagements


"@NoahEpstein_ Hey Meta can try but I promise you they're not going to get it. Llama is horrible"  
[X Link](https://x.com/jackhnels/status/2022535142837731518)  2026-02-14T04:55Z [--] followers, [---] engagements


"@jwsaml Hey if you haven't tried it yet give Minimax M2.5 a try as well as GLM5. Both are incredibly cheap way cheaper than GPT [---] and are scoring at nearly the same level as Opus 4.6"  
[X Link](https://x.com/jackhnels/status/2022536042490868211)  2026-02-14T04:59Z [--] followers, [--] engagements


"@BrianRoemmele How are you currently doing it with NATO transactions through X402 or something else"  
[X Link](https://x.com/jackhnels/status/2022536502086083025)  2026-02-14T05:00Z [--] followers, [---] engagements


"@Jall_n How the heck have I not heard of Yahoo Add MCP and Stripe shipping X402 payments"  
[X Link](https://x.com/jackhnels/status/2022536829598138846)  2026-02-14T05:02Z [--] followers, [---] engagements


"@GanimCorey @openclaw Have you ever used Claude code to debug Openclaw to debug Claude code Because I have I can tell you that's another bigger unlock"  
[X Link](https://x.com/jackhnels/status/2022537168443277742)  2026-02-14T05:03Z [--] followers, [--] engagements


"@mikeejt @iyoushetwt Did you see their update yesterday"  
[X Link](https://x.com/jackhnels/status/2022539204648227031)  2026-02-14T05:11Z [--] followers, [---] engagements


"@OpenDeMind @NoahEpstein_ @ManusAI Dang maybe Ill give it another shot Im just a massive openclaw fan"  
[X Link](https://x.com/jackhnels/status/2022673131786195009)  2026-02-14T14:03Z [--] followers, [--] engagements


"@KyleRaap @openclaw I haven't had me ask many questions but I definitely agree that Opus should be the orchestrator and Minimax K2.5 should be the sub-agents"  
[X Link](https://x.com/jackhnels/status/2022741640692088841)  2026-02-14T18:36Z [--] followers, [--] engagements


"@openclaw If you struggle with token costs for Opus [---] this is the update for you. Minimax M2.5 is the go-to for a cheaper alternative to Opus [---] this paired with Kimi K2.5 and there's no reason to be spending $15 per million tokens"  
[X Link](https://x.com/jackhnels/status/2022137125207257200)  2026-02-13T02:33Z [--] followers, 12.4K engagements


"Minimax M [---] is out for Openclaw. Vibe coding benchmarks have it just [---] points away from Opus [---]. That's higher than GPT-5.2 Codex higher than Kimi K2.5 and higher than GLM. If you're sick of spending thousands on tokens every month this is the way to go"  
[X Link](https://x.com/anyuser/status/2022139578124251288)  2026-02-13T02:43Z [--] followers, [---] engagements


"@Null1thm @openclaw I pay $300 a month between chatgpt pro and claude max. Trust me this is the way"  
[X Link](https://x.com/jackhnels/status/2022250278448488719)  2026-02-13T10:03Z [--] followers, [---] engagements


"@BastianelliLore @openclaw The benchmarks are showing its performing better with vibe coding tasks and potentially real world results. Test for yourself GLM performed better on the benchmarks and actually sucks in real world application"  
[X Link](https://x.com/jackhnels/status/2022250468596977683)  2026-02-13T10:04Z [--] followers, [---] engagements


"If you use Openclaw pay attention. For weeks you've heard that Opus [---] is the best model. The game has just changed forever. Mini Max M [---] came out yesterday with native use for Openclaw. Say goodbye to All of your $200+ subscriptions and API costs: Tool calling (BFCL): MiniMax [----] vs Opus [--] vs GPT-5.2 [--] Bug fixing (SWE-Bench): [----] vs [----] vs [--] Multi-repo fixes: [----] vs [----] vs [----] Now look at the price per million tokens: MiniMax M2: $0.27 in / $0.95 out Opus 4.6: $5.00 in / $25.00 out GPT-5.2: $1.75 in / $14.00 out MiniMax is 18x cheaper than Opus on input. 26x cheaper on output."  
[X Link](https://x.com/anyuser/status/2022255720729567367)  2026-02-13T10:25Z [--] followers, [---] engagements


"@jakub_rusiecki This is what I've been telling everybody. Openclaw was just V1 the first company to build a UI with no-code solutions where your typical person can just create an account and start building. That's the software that's going to win it all"  
[X Link](https://x.com/jackhnels/status/2022274056528023807)  2026-02-13T11:38Z [--] followers, [--] engagements


"Recursive language models for those who don't understand are basically an orchestrator LLM that uses code to section out a large document or file anything above its context window then it deploys many LLMs to digest a portion of that large file and feed it back to the orchestrator LLM. Think of it how a CEO can only get done so much as one person but if he hires employees he can get a lot more done. Those employees report back to him. If the employees don't give him enough detail he can send the employees out to do more research which is exactly how recursive Language models work"  
[X Link](https://x.com/jackhnels/status/2022274723183313003)  2026-02-13T11:40Z [--] followers, [---] engagements


"Ive tested all three extensively. Heres what Ive learned: Opus [---] - the gold standard. Perfect for every task. Coding chatting thinking everything Kimi K2.5 - the cheap one. Kimi loops over and over on itself. Its coding intelligence sucks. Its not great at chatting. Thinking is pretty good. Minimax M2.5 - cheapest one. Doesnt loop on itself. Coding is great not as good as opus but veeery close. Thinking is very good not as good as opus but again its close. Tool calls are better here. Context is smaller but more efficiently used. Overall Opus is the best BUT Minimax is close. For being about"  
[X Link](https://x.com/jackhnels/status/2022384434289479724)  2026-02-13T18:56Z [--] followers, [----] engagements


"2 MASSIVE updates in this one for those who missed it: [--]. GPT-5.3-Codex-Spark support This means you now how have THE quickest coding model out there right now which also goes head to head against Opus [---]. (cheaper too) 1k+tokens/second [--]. Threading that ACTUALLY works. This means that every conversation you have with your Openclaw improves upon the last. No more amnesia. On top of this you also get concurrent processing of various data streams. In simple words Pete just added another model that pushes the bounds of what's possible with AI and allows models to actually improve over time"  
[X Link](https://x.com/jackhnels/status/2022532920716448120)  2026-02-14T04:46Z [--] followers, [----] engagements


"@JeffKirdeikis This is what I've been preaching from the rooftops. Goldman Sachs is already replacing their backroom employees"  
[X Link](https://x.com/jackhnels/status/2022536671997378884)  2026-02-14T05:01Z [--] followers, [---] engagements


"@ericosiu What are you using for the Open Graph UI I've seen some good things come up recently with rowboat and other sorts"  
[X Link](https://x.com/jackhnels/status/2022830996249321640)  2026-02-15T00:31Z [--] followers, [--] engagements


"@nummanali @Kimi_Moonshot @MiniMax_AI @Zai_org @OpenRouterAI @vercel @ollama @NanoGPTcom @GatewayzAI This seems like a really cool idea. What's the purpose of using this though"  
[X Link](https://x.com/jackhnels/status/2022832953793933355)  2026-02-15T00:38Z [--] followers, [---] engagements


"@Jush21e8 Why do you say that Seems like you haven't used it enough. The first couple of hours of use suck but after that it starts to quickly pick up"  
[X Link](https://x.com/jackhnels/status/2022833221298221258)  2026-02-15T00:39Z [--] followers, [----] engagements


"@sillydarket Excited to see an improvement in agentic memory How are you building this What's the end goal"  
[X Link](https://x.com/jackhnels/status/2022833375883444514)  2026-02-15T00:40Z [--] followers, [---] engagements


"@eachlabs Does it have an API I can use for my Openclaw"  
[X Link](https://x.com/jackhnels/status/2022834290728325567)  2026-02-15T00:44Z [--] followers, [--] engagements


"@Aurelien_Gz This feels like vibe coding but less efficient"  
[X Link](https://x.com/jackhnels/status/2022835639918481565)  2026-02-15T00:49Z [--] followers, [--] engagements


"@santisiri @steipete @openclaw He's a developer not a business man. It makes sense why he's against profit"  
[X Link](https://x.com/jackhnels/status/2022836085135421532)  2026-02-15T00:51Z [--] followers, [--] engagements


"This is super helpful for every agent. Is it open source or does it cost something to use and implement I'd love to get working on this and implement it into my own workflow. I'd be happy to test it and let you know what I find I've been doing a ton of memory and context graph for my agents as well and trying to build systems https://twitter.com/i/web/status/2022837427040117077 https://twitter.com/i/web/status/2022837427040117077"  
[X Link](https://x.com/jackhnels/status/2022837427040117077)  2026-02-15T00:56Z [--] followers, [--] engagements


"@nummanali @Kimi_Moonshot @MiniMax_AI @Zai_org @OpenRouterAI @vercel @ollama @NanoGPTcom @GatewayzAI Very interesting. What makes quad code a better harness than something like Openclaw for coding"  
[X Link](https://x.com/jackhnels/status/2022837555935269244)  2026-02-15T00:57Z [--] followers, [--] engagements


"@LLMJunky @nummanali @Kimi_Moonshot @MiniMax_AI @Zai_org @OpenRouterAI @vercel @ollama @NanoGPTcom @GatewayzAI I can definitely see why that'd be difficult. I guess I just don't understand why Clawdbot Code is the premier way to code when things like Openclaw exist"  
[X Link](https://x.com/jackhnels/status/2022837670167146544)  2026-02-15T00:57Z [--] followers, [--] engagements


"@adriansolarzz Hey if it works for you it works for you I can definitely see this being successful if you replace many chat with it but personally just not for me"  
[X Link](https://x.com/jackhnels/status/2022837898966409561)  2026-02-15T00:58Z [--] followers, [--] engagements


"@LLMJunky @nummanali @Kimi_Moonshot @MiniMax_AI @Zai_org @OpenRouterAI @vercel @ollama @NanoGPTcom @GatewayzAI Understood"  
[X Link](https://x.com/jackhnels/status/2022913230285500497)  2026-02-15T05:57Z [--] followers, [--] engagements


"BREAKING: Kimi is releasing Kimi Claw an OpenClaw wrapper native in Kimi through the browser. With Manus coming out just a few days ago online 24/7 agents are becoming a new trend. Keep an eye on this Introducing Kimi Claw🦞 OpenClaw now native to https://t.co/8PH4O9dh6m. Living right in your browser tab online 24/7. 🔹 ClawHub Access: 5000+ community skills in the ClawHub library. 🔹 40GB Cloud Storage: Massive space for all your files 🔹 Pro-Grade Search: Fetch live https://t.co/eWwp6sWEw3 Introducing Kimi Claw🦞 OpenClaw now native to https://t.co/8PH4O9dh6m. Living right in your browser"  
[X Link](https://x.com/anyuser/status/2023043857034965066)  2026-02-15T14:36Z [--] followers, [--] engagements


"Introducing Kimi Claw🦞 OpenClaw now native to Living right in your browser tab online 24/7. 🔹 ClawHub Access: 5000+ community skills in the ClawHub library. 🔹 40GB Cloud Storage: Massive space for all your files 🔹 Pro-Grade Search: Fetch live high-quality data directly from Yahoo Finance and more. 🔹 Bring Your Own Claw: Connect your third-party OpenClaw to chat with your setup or bridge it to apps like Telegram groups. Discover call and chain them instantly within Beta Access: Now open for Allegretto members and above. Try it now at: https://www.kimi.com/bot http://kimi.com"  
[X Link](https://x.com/anyuser/status/2023029674549596301)  2026-02-15T13:40Z 106.7K followers, 157.4K engagements


"If you use Openclaw pay attention. For weeks you've heard that Opus [---] is the best model. The game has just changed forever. Mini Max M [---] came out yesterday with native use for Openclaw. Say goodbye to All of your $200+ subscriptions and API costs: Tool calling (BFCL): MiniMax [----] vs Opus [--] vs GPT-5.2 [--] Bug fixing (SWE-Bench): [----] vs [----] vs [--] Multi-repo fixes: [----] vs [----] vs [----] Now look at the price per million tokens: MiniMax M2: $0.27 in / $0.95 out Opus 4.6: $5.00 in / $25.00 out GPT-5.2: $1.75 in / $14.00 out MiniMax is 18x cheaper than Opus on input. 26x cheaper on output."  
[X Link](https://x.com/anyuser/status/2022255720729567367)  2026-02-13T10:25Z [--] followers, [---] engagements


"Minimax M [---] is out for Openclaw. Vibe coding benchmarks have it just [---] points away from Opus [---]. That's higher than GPT-5.2 Codex higher than Kimi K2.5 and higher than GLM. If you're sick of spending thousands on tokens every month this is the way to go"  
[X Link](https://x.com/anyuser/status/2022139578124251288)  2026-02-13T02:43Z [--] followers, [---] engagements


"If Opus [---] is the only model you're using for your Openclaw what are you even doing With Opus [---] you have two options: [--]. Get limited even with a $200/month plan. [--]. Spend thousands a month on Anthropic API This is the part where I urge you to explore other models especially ones that can be run on a 3rd party API provider https://twitter.com/i/web/status/2021995483216920706 https://twitter.com/i/web/status/2021995483216920706"  
[X Link](https://x.com/anyuser/status/2021995483216920706)  2026-02-12T17:11Z [--] followers, [---] engagements


"The future of software and tech will belong to people who: - act quick - learn quick - adapt quick - explain things simply Everyone else will get lost in the weeds trying to think it through. Moral of the story ACT. Get your feet wet with every new software and AI advancement today"  
[X Link](https://x.com/anyuser/status/2021900848817643710)  2026-02-12T10:55Z [--] followers, [---] engagements


"Openclaw BURNS through your claude tokens like crazy. Heres how Im saving thousands on token costs now: Codex CLI - $1.50/M input $6/M output. Runs locally uses your existing ChatGPT subscription and handles coding tasks that would eat through opus usage. Kimi K2.5 - $0.60/M input $3/M output. [--] trillion parameters but only activates 32B per query (MoE architecture). 100% open source too. Benchmarks competitive with models 10x the price. For context: Claude Opus is $15/$75 per million tokens. You don't need the most expensive model for every task. Route simple stuff to cheaper models save the"  
[X Link](https://x.com/anyuser/status/2021897441323168176)  2026-02-12T10:41Z [--] followers, [---] engagements


"Big Tech is spending $650 billion on AI in [----]. Here's who's writing the biggest checks: Amazon: $200B (AWS infrastructure) Alphabet: $175-185B (Gemini Google Cloud) Microsoft: $145B (Azure OpenAI Copilot) Meta: $115-135B (Llama models AI ads) That's roughly 15-20% of their combined revenue. For context these four companies pull in over $1.5T/year. The bet isn't speculative anymore. Meta just crossed $200B in annual revenue driven almost entirely by AI-powered ad targeting. The ROI is real. The risk isn't overspending. It's underspending and getting disrupted. And here's a reusable tracking"  
[X Link](https://x.com/anyuser/status/2021207980813713537)  2026-02-10T13:01Z [--] followers, [--] engagements


"Everyone's talking about the OpenAI vs Anthropic Super Bowl ads. Here's what actually matters: both companies just released new flagship models the same week and nobody's talking about that. Anthropic dropped Opus [---]. OpenAI shipped GPT-5.3-Codex with a full agent platform called Frontier. The ad war is theater. The agent war is real. And the company that wins the "build stuff without code" market wins the next decade of business software. Who do you think is winning"  
[X Link](https://x.com/anyuser/status/2021173663626613040)  2026-02-10T10:45Z [--] followers, [--] engagements


"Running Kimi K2 locally costs more than most people's car payment. [--] trillion parameters. 500GB VRAM at INT4. That's 7x RTX 5090s at $3000+ each. $21000 in GPUs alone. And with 4000W under load thats around $350/month in electricity just to keep it running. Or you pay $0.15 per million input tokens on a pay-per-use API platform and get the same model. Streaming. No setup. No heat. No driver issues. A million tokens through the API costs less than a gas station coffee. Stop romanticizing local inference"  
[X Link](https://x.com/anyuser/status/2020865921393246713)  2026-02-09T14:22Z [--] followers, [--] engagements


"5 things most people get wrong with OpenClaw: [--]. They skip SOUL md and MEMORY md. Your agent wakes up with amnesia every session. These files ARE its memory. Set them up or you're talking to a stranger every time. [--]. They run everything on Opus. Model routing exists for a reason. Haiku for cron jobs. Sonnet for sub-agents. Opus for the stuff that actually matters. Your API bill will thank you. [--]. They do everything in the main session. Sub-agents exist. Spawn them. Let them do the heavy lifting in the background while you keep working. [--]. They never set up cron jobs. OpenClaw can check your"  
[X Link](https://x.com/anyuser/status/2020815775020335437)  2026-02-09T11:03Z [--] followers, [--] engagements


"What are your honest thoughts on @aidotcom I heard they were requiring a card on file just to verify that users are human"  
[X Link](https://x.com/anyuser/status/2020685623951310874)  2026-02-09T02:26Z [--] followers, [--] engagements


"OpenClaw was drowning my tokens. Burning $50/day on API calls. Forgetting everything between sessions. I thought the tool was broken. Turns out I was building it wrong. Here's how I fixed it 👇🧵"  
[X Link](https://x.com/anyuser/status/2020683802738405549)  2026-02-09T02:18Z [--] followers, [--] engagements


"Running [--] cron jobs daily. Each one spawned a new session. [----] stale sessions piling up in memory. Gateway scanning all of them on every request. Latency got worse. Token spend got worse. No persistent memory anywhere. Three changes: [--]. Killed the junk cron jobs (GChat indexing weekly content batches Friday loop closers) [--]. Aggressive session cleanup: run every [--] min nuke anything [--] hours old [--]. Built actual memory: mem0 + ChromaDB for persistent recall across sessions Now Token spend cut by 60%. Response time halved. Agent actually remembers context from last week. The lesson: frameworks"  
[X Link](https://x.com/anyuser/status/2020684226975179165)  2026-02-09T02:20Z [--] followers, [--] engagements


"# Prompt: Write a Technical Guide Post Write a beginner-friendly technical guide post on installing and setting up mem0 + ChromaDB + Llama for RAG and memory systems. Target audience: developers interested in local AI agents and long-term memory for LLM applications. ## Include These Sections [--]. **Intro** Why you'd want this stack (local control privacy persistent memory for agents) [--]. **Prerequisites** What you need (Python 3.9+ 8GB+ RAM Homebrew/apt basic CLI comfort) [--]. **Installation Steps** (in order): - Install Ollama and pull Llama model - Install ChromaDB (pip install chromadb) -"  
[X Link](https://x.com/anyuser/status/2020684306847244692)  2026-02-09T02:20Z [--] followers, [--] engagements


"My AI agent was burning tokens like a college kid with a new credit card. Ran a full audit today. Cut API costs 67% in one session. Here's exactly what was wrong and how I fixed it"  
[X Link](https://x.com/anyuser/status/2020600585741803618)  2026-02-08T20:48Z [--] followers, [--] engagements


"Fix 5: cron prompt diet. Every token in a cron prompt gets paid on every run. A job running 4x daily with a [----] token prompt = [----] wasted input tokens per day. Shortened all [--] cron prompts by 50-80%. Same instructions fewer words"  
[X Link](https://x.com/anyuser/status/2020601024424362027)  2026-02-08T20:49Z [--] followers, [--] engagements


"Put it all together: Context per prompt: -67% Heartbeat cost: -67% Cron job cost: -50-80% Rate limit failures: 0% Freed 560MB RAM (stopped unused Docker containers) Set up a weekly audit cron to catch bloat automatically. Runs every Saturday. Full guide with config examples checklists and exact settings: Free. No gates. Copy-paste into your setup. http://github.com/Jacknelson6/openclaw-optimization-guide http://github.com/Jacknelson6/openclaw-optimization-guide"  
[X Link](https://x.com/anyuser/status/2020601214669385746)  2026-02-08T20:50Z [--] followers, [--] engagements

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