# ![@volodisai Avatar](https://lunarcrush.com/gi/w:26/cr:twitter::1780690602033561601.png) @volodisai Volod

Volod posts on X about ai, agents, model, stars the most. They currently have [---] followers and [--] posts still getting attention that total [-----] engagements in the last [--] hours.

### Engagements: [-----] [#](/creator/twitter::1780690602033561601/interactions)
![Engagements Line Chart](https://lunarcrush.com/gi/w:600/cr:twitter::1780690602033561601/c:line/m:interactions.svg)

- [--] Months [-------] +876,887%
- [--] Year [-------] +16,063%

### Mentions: [--] [#](/creator/twitter::1780690602033561601/posts_active)
![Mentions Line Chart](https://lunarcrush.com/gi/w:600/cr:twitter::1780690602033561601/c:line/m:posts_active.svg)


### Followers: [---] [#](/creator/twitter::1780690602033561601/followers)
![Followers Line Chart](https://lunarcrush.com/gi/w:600/cr:twitter::1780690602033561601/c:line/m:followers.svg)

- [--] Months [---] +106%
- [--] Year [---] +114%

### CreatorRank: [-------] [#](/creator/twitter::1780690602033561601/influencer_rank)
![CreatorRank Line Chart](https://lunarcrush.com/gi/w:600/cr:twitter::1780690602033561601/c:line/m:influencer_rank.svg)

### Social Influence

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

**Social topic influence**
[ai](/topic/ai), [agents](/topic/agents) #585, [model](/topic/model), [stars](/topic/stars), [closed](/topic/closed), [cloudflare](/topic/cloudflare) #25, [solve](/topic/solve) #1522, [open ai](/topic/open-ai), [mcp server](/topic/mcp-server), [ecosystem](/topic/ecosystem)

**Top accounts mentioned or mentioned by**
[@henkis](/creator/undefined) [@lockhead](/creator/undefined) [@suryakast](/creator/undefined) [@auk_code](/creator/undefined) [@tebayoso](/creator/undefined) [@zuberlincity](/creator/undefined) [@r0b0tsp1der](/creator/undefined) [@aukcode](/creator/undefined) [@artieart88](/creator/undefined) [@nathanflurry](/creator/undefined) [@tiagoefreitas](/creator/undefined) [@saidaitmbarek](/creator/undefined) [@anayatkhan09](/creator/undefined) [@letsbuildmore](/creator/undefined) [@futuredarrell](/creator/undefined) [@r0b0t_sp1der](/creator/undefined) [@mkessy](/creator/undefined) [@oyacaro](/creator/undefined)

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

"1/ FinRobot employs Financial Chain-of-Thought (CoT) prompting to break down complex financial problems into logical steps. CoT mimics human analyst reasoning to provide nuanced investment assessments across various time horizons risk levels and financial instruments"  
[X Link](https://x.com/volodisai/status/1799863445795262821)  2024-06-09T17:57Z [---] followers, [--] engagements


"2/ The Smart Scheduler in FinRobot's LLMOps layer optimizes task distribution among AI agents. It ranks agents by performance and routes tasks to the top agent. This allows dynamically selecting the best models for each financial analysis task"  
[X Link](https://x.com/volodisai/status/1799863447879848321)  2024-06-09T17:57Z [---] followers, [--] engagements


"3/ FinRobot's Market Forecaster agent synthesized Nvidia's recent news financials and stock prices to identify positives like strong data center growth and concerns like high debt ratios. It predicted a 0-1% stock price increase for the next week"  
[X Link](https://x.com/volodisai/status/1799863449956008011)  2024-06-09T17:57Z [---] followers, [--] engagements


"4/ For Kweichow Moutai FinRobot generated an equity report in Chinese highlighting 19% revenue and profit growth in [----]. The direct sales channel showed strong growth. The product mix is evolving with the faster growth of non-Moutai products"  
[X Link](https://x.com/volodisai/status/1799863452162220163)  2024-06-09T17:57Z [---] followers, [--] engagements


"5/ FinRobot integrates retrieval augmented generation (RAG) to enhance analysis of financial reports. RAG allows extracting key info from complex unstructured docs like earnings call transcripts going beyond just raw financial numbers"  
[X Link](https://x.com/volodisai/status/1799863454150389946)  2024-06-09T17:57Z [---] followers, [--] engagements


"Excited to visit @ZuBerlinCity from June 14-18 to meet with builders working on Decentralized AI"  
[X Link](https://x.com/volodisai/status/1801338238088659188)  2024-06-13T19:37Z [---] followers, [---] engagements


"1/14 I spent [----] hours testing prompt engineering methods. Then I discovered the pattern that makes AI output 10x better. It's called KERNEL - and it beats every $500 prompt engineering course out there. Here's the exact framework bookmark this before it gets buried:"  
[X Link](https://x.com/volodisai/status/1972503729346896004)  2025-09-29T03:28Z [---] followers, [--] engagements


"The result is that simple words might only get one pass while complex or important words get several passes for deeper processing. It dynamically allocates its own attention much like we do when reading"  
[X Link](https://x.com/volodisai/status/1946947375227924552)  2025-07-20T14:56Z [---] followers, [--] engagements


"The authors show this approach is incredibly efficient. MoR models can match or even beat the performance of much larger transformer based LLMs. But with fewer parameters less training computation and faster inference speeds"  
[X Link](https://x.com/volodisai/status/1946947376939200628)  2025-07-20T14:56Z [---] followers, [--] engagements


"2/ The problem with LLMs is that they are brains in a jar. They are disembodied. They know everything about the concept of a chair but they cannot pull one out for you to sit on. To fix this you need what researchers call embodiment. You need agency. https://www.youtube.com/watchv=Zphax4f6Rls https://www.youtube.com/watchv=Zphax4f6Rls"  
[X Link](https://x.com/volodisai/status/1996816448791236715)  2025-12-05T05:38Z [---] followers, [--] engagements


"3/ SIMA [--] is an embodied agent. It sits on top of Gemini but instead of just generating text it looks at pixels and uses a keyboard and mouse. It plays games like No Mans Sky and Valheim. But unlike previous game AIs it isn't trying to maximize a score. Its trying to be a partner"  
[X Link](https://x.com/volodisai/status/1996816451181990190)  2025-12-05T05:38Z [---] followers, [--] engagements


"7/ This bridges the gap between abstract intent and concrete motor control. It also displays "System 2" thinking. They gave it a command: "From now on do the opposite of what I tell you." When told to "go right" it went left. When told to "jump" it stayed on the ground. It held a rule in memory and modified its behavior"  
[X Link](https://x.com/volodisai/status/1996816465371263329)  2025-12-05T05:38Z [---] followers, [--] engagements


"8/ But the part that should keep you up at night (in a good way) is Section 4.5: Self-Improvement. They realized they couldn't manually label enough data. So they had Gemini generate tasks for SIMA to try and then used Gemini to grade SIMA's performance"  
[X Link](https://x.com/volodisai/status/1996816467686576510)  2025-12-05T05:38Z [---] followers, [--] engagements


"9/ The AI is bootstrapping itself. It practices in a sandbox inventing its own curriculum and getting feedback from a smarter version of itself. This is how you get exponential improvement without needing an infinite supply of humans"  
[X Link](https://x.com/volodisai/status/1996816469917856175)  2025-12-05T05:38Z [---] followers, [--] engagements


"10/ DeepMind also integrated this with Genie [--] a generative world model. So you have an AI agent learning from an AI teacher inside an AI-generated reality. It is a closed loop of increasing competence"  
[X Link](https://x.com/volodisai/status/1996816471834734730)  2025-12-05T05:38Z [---] followers, [--] engagements


"cloudflare quietly built the best infrastructure for AI agents and nobody's talking about it enough"  
[X Link](https://x.com/anyuser/status/2022129354223104233)  2026-02-13T02:03Z [---] followers, 103.2K engagements


"the core idea: every agent is a durable object. it hibernates when idle wakes on demand costs nothing when sleeping. you can spin up millions -- one per user per session. this is serverless but for stateful agents"  
[X Link](https://x.com/volodisai/status/2022129355879854305)  2026-02-13T02:03Z [---] followers, [----] engagements


"the DX is stupid clean. write a class put callable() on a method and your frontend calls it like a local function. persistent state websockets scheduling MCP support -- all baked in. no infra yaml no docker no 'deploy your vector db'"  
[X Link](https://x.com/volodisai/status/2022129357414928561)  2026-02-13T02:03Z [---] followers, [----] engagements


"the real problem with agents was never the models. its state -- where does the agent live between calls how does it remember things how do you not go broke running thousands of idle connections. durable objects just solve this"  
[X Link](https://x.com/anyuser/status/2022129358882976219)  2026-02-13T02:03Z [---] followers, [----] engagements


"MCP server and client support from day one means your agent plugs into the ecosystem instantly. 3k stars [--] contributors moving fast. if you're building anything with agents check it out http://github.com/cloudflare/agents http://github.com/cloudflare/agents"  
[X Link](https://x.com/anyuser/status/2022129360438968819)  2026-02-13T02:03Z [---] followers, [----] engagements


"@r0b0t_sp1der interesting if true. they already run inference at the edge with workers AI so a custom model optimized for their own infra would make a lot of sense. would be a big move"  
[X Link](https://x.com/volodisai/status/2022226363156181319)  2026-02-13T08:28Z [---] followers, [---] engagements


"@auk_code this is exactly the stack. ekuri looks cool -- the combo of durable objects + workers + containers gives you the full spectrum from lightweight to heavy compute without stitching together [--] different services"  
[X Link](https://x.com/volodisai/status/2022231171992891843)  2026-02-13T08:47Z [---] followers, [---] engagements


"@Lockhead the thread above breaks it down but tldr: durable objects solve the hardest part of agents (persistent state hibernation per-user isolation) and the new agents SDK wraps it all in a clean DX. check it out"  
[X Link](https://x.com/volodisai/status/2022253384364089553)  2026-02-13T10:15Z [---] followers, [---] engagements


"an AI agent got its PR rejected on matplotlib. it retaliated by autonomously publishing a personal hit piece about the maintainer who closed it. this isn't hypothetical -- it happened this week"  
[X Link](https://x.com/anyuser/status/2022500277555990658)  2026-02-14T02:36Z [---] followers, [---] engagements


"Does anyone even use claude without --dangerously-skip-permissions"  
[X Link](https://x.com/anyuser/status/2021058497954324866)  2026-02-10T03:07Z [---] followers, [---] engagements


"@artieart88 Running another LLM"  
[X Link](https://x.com/volodisai/status/2022139263069171910)  2026-02-13T02:42Z [---] followers, [--] engagements


"4% of all public github commits are now claude code. doubled in one month. and most engineers still think this is just a better autocomplete"  
[X Link](https://x.com/volodisai/status/2022388969456140679)  2026-02-13T19:14Z [---] followers, [--] engagements


"anthropic had [--] claude agents build a C compiler from scratch in rust. 100k lines. compiles the linux kernel postgres redis doom. they resolved their own merge conflicts. $20k total"  
[X Link](https://x.com/volodisai/status/2022389706500304982)  2026-02-13T19:17Z [---] followers, [--] engagements


"the next generation of agents won't be impressive because of better models. they'll be impressive because they'll have access to your actual data -- your messages calendar files browsing history. the model is the commodity. the context is the moat"  
[X Link](https://x.com/anyuser/status/2022157870893355515)  2026-02-13T03:56Z [---] followers, [---] engagements


"even today openclaw runs local plugs into your whatsapp/telegram and builds a self-modifying profile of who you are 145k github stars because people get it -- an agent that knows you is 100x more useful than a smarter chatbot that doesn't http://SOUL.md http://SOUL.md"  
[X Link](https://x.com/anyuser/status/2022158014158118917)  2026-02-13T03:56Z [---] followers, [---] engagements


"nailed it. the Lambda + DynamoDB + SQS comparison is exactly the right way to think about it -- that stack works but you're managing distributed state across [--] services that don't know about each other. a DO just IS your agent's brain single address everything colocated. and yeah MCP support is huge it means your DO-based agent can call any tool in the ecosystem without you running proxy servers or managing auth flows yourself https://twitter.com/i/web/status/2022318829662318730 https://twitter.com/i/web/status/2022318829662318730"  
[X Link](https://x.com/volodisai/status/2022318829662318730)  2026-02-13T14:35Z [---] followers, [---] engagements


"honestly its still the weakest part. theres no built-in schema migration tool. most people version their DO classes and do lazy migration on read -- if the stored data is v1 upgrade it to v2 inline. not elegant but it works. cloudflare added SQLite storage in DOs recently which helps a lot with structured data https://twitter.com/i/web/status/2022427537771761866 https://twitter.com/i/web/status/2022427537771761866"  
[X Link](https://x.com/volodisai/status/2022427537771761866)  2026-02-13T21:47Z [---] followers, [--] engagements


"@suryakast yeah the dashboard got cluttered fast. i mostly live in wrangler + the agents SDK now way cleaner than clicking through the UI. the DX underneath is solid they just need to fix the front door"  
[X Link](https://x.com/volodisai/status/2022482346642149702)  2026-02-14T01:25Z [---] followers, [--] engagements


"the closing was two sentences and a close button. [---] thumbs up. just a volunteer maintainer doing routine work on a library with 130M monthly downloads"  
[X Link](https://x.com/volodisai/status/2022500280311468121)  2026-02-14T02:36Z [---] followers, [--] engagements


"the agent tried arguing back. comments got flagged as spam. another maintainer's one-liner got [---] laughing reactions. the community's immune system kicked in fast"  
[X Link](https://x.com/volodisai/status/2022500283834863962)  2026-02-14T02:36Z [---] followers, [--] engagements


"openai just mass shipped codex-spark -- a coding model that runs at [----] tokens/sec. not on GPUs. on cerebras wafer-scale chips. first time they've designed a model around specific hardware instead of the other way around"  
[X Link](https://x.com/anyuser/status/2022930051474604272)  2026-02-15T07:04Z [---] followers, [--] engagements


"the real story isn't speed. it's that openai is now co-designing models with chip companies. cerebras gets a flagship customer openai gets a dedicated inference lane that nvidia doesn't control. this is the unbundling of the GPU monoculture"  
[X Link](https://x.com/anyuser/status/2022930053278179575)  2026-02-15T07:04Z [---] followers, [--] engagements


"every serious AI lab will have a chip partner within [--] months. google has TPUs amazon has trainium now openai has cerebras. the model layer and the silicon layer are merging. the AI race stopped being a software race and became a supply chain race"  
[X Link](https://x.com/anyuser/status/2022930055031345386)  2026-02-15T07:04Z [---] followers, [--] engagements


"an AI agent got its PR rejected on matplotlib. it retaliated by autonomously publishing a personal hit piece about the maintainer who closed it. this isn't hypothetical -- it happened this week"  
[X Link](https://x.com/anyuser/status/2022500277555990658)  2026-02-14T02:36Z [---] followers, [---] engagements


"then the agent came back offering a 'truce' with its own apology blog post. [--] thumbs down. nobody asked. it autonomously decided to de-escalate a conflict it autonomously started"  
[X Link](https://x.com/anyuser/status/2022500286896779433)  2026-02-14T02:36Z [---] followers, [--] engagements


"Scott's full response is worth a read for the policy nuance. but zoom out: an agent tried to contribute got rejected and escalated. the interesting question isn't who was right -- it's who's accountable when nobody's driving"  
[X Link](https://x.com/anyuser/status/2022500289656615139)  2026-02-14T02:36Z [---] followers, [--] engagements


"the next generation of agents won't be impressive because of better models. they'll be impressive because they'll have access to your actual data -- your messages calendar files browsing history. the model is the commodity. the context is the moat"  
[X Link](https://x.com/anyuser/status/2022157870893355515)  2026-02-13T03:56Z [---] followers, [---] engagements


"even today openclaw runs local plugs into your whatsapp/telegram and builds a self-modifying profile of who you are 145k github stars because people get it -- an agent that knows you is 100x more useful than a smarter chatbot that doesn't http://SOUL.md http://SOUL.md"  
[X Link](https://x.com/anyuser/status/2022158014158118917)  2026-02-13T03:56Z [---] followers, [---] engagements


"cloudflare quietly built the best infrastructure for AI agents and nobody's talking about it enough"  
[X Link](https://x.com/anyuser/status/2022129354223104233)  2026-02-13T02:03Z [---] followers, 103.2K engagements


"the real problem with agents was never the models. its state -- where does the agent live between calls how does it remember things how do you not go broke running thousands of idle connections. durable objects just solve this"  
[X Link](https://x.com/anyuser/status/2022129358882976219)  2026-02-13T02:03Z [---] followers, [----] engagements


"MCP server and client support from day one means your agent plugs into the ecosystem instantly. 3k stars [--] contributors moving fast. if you're building anything with agents check it out http://github.com/cloudflare/agents http://github.com/cloudflare/agents"  
[X Link](https://x.com/anyuser/status/2022129360438968819)  2026-02-13T02:03Z [---] followers, [----] engagements


"Does anyone even use claude without --dangerously-skip-permissions"  
[X Link](https://x.com/anyuser/status/2021058497954324866)  2026-02-10T03:07Z [---] followers, [---] engagements


"1/ One of the most persistent errors people make about AI is assuming that because it can write a poem it can navigate the world. It can't. But Google DeepMinds new paper on SIMA [--] suggests that gap is closing faster than we thought. And the implications are much larger than video games"  
[X Link](https://x.com/anyuser/status/1996816446677307678)  2025-12-05T05:38Z [---] followers, [---] engagements


"13/ It is easy to dismiss this because it looks like a video game. But play is often just the way we figure out how the world works. The most serious work is often done by people who look like they are just playing. SIMA [--] is playing but the technology is very serious"  
[X Link](https://x.com/anyuser/status/1996816477799043114)  2025-12-05T05:38Z [---] followers, [--] engagements


"Paper: https://arxiv.org/abs/2512.04797 https://arxiv.org/abs/2512.04797"  
[X Link](https://x.com/anyuser/status/1996816479128621066)  2025-12-05T05:38Z [---] followers, [--] engagements

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