[GUEST ACCESS MODE: Data is scrambled or limited to provide examples. Make requests using your API key to unlock full data. Check https://lunarcrush.ai/auth for authentication information.] #  @Dan_Jeffries1 Daniel Jeffries Daniel Jeffries posts on X about faster, its just, build a, day of the most. They currently have XXXXXX followers and XXX posts still getting attention that total XXXXX engagements in the last XX hours. ### Engagements: XXXXX [#](/creator/twitter::633549366/interactions)  - X Week XXXXXX -XX% - X Month XXXXXXX -XX% - X Months XXXXXXXXX -XX% - X Year XXXXXXXXX -XX% ### Mentions: X [#](/creator/twitter::633549366/posts_active)  - X Year XXX -XXXX% ### Followers: XXXXXX [#](/creator/twitter::633549366/followers)  - X Week XXXXXX -XXXX% - X Month XXXXXX +0.14% - X Months XXXXXX +7.90% - X Year XXXXXX +23% ### CreatorRank: XXXXXXX [#](/creator/twitter::633549366/influencer_rank)  ### Social Influence [#](/creator/twitter::633549366/influence) --- **Social category influence** [social networks](/list/social-networks) [technology brands](/list/technology-brands) **Social topic influence** [faster](/topic/faster), [its just](/topic/its-just), [build a](/topic/build-a), [day of](/topic/day-of), [a very](/topic/a-very), [1m](/topic/1m), [token](/topic/token), [youtube](/topic/youtube), [over the](/topic/over-the), [all the](/topic/all-the) ### Top Social Posts [#](/creator/twitter::633549366/posts) --- Top posts by engagements in the last XX hours "At what point does the cognitive dissonance kick and you realize your AI Doomer movement is a joke Is it when the 'If You Build It Everyone Dies' falls off the best seller list in a week and you realize your astroturfing foundations that bought enough copies to put it there for that one week are the only copies getting sold before it ends up at Goodwill for XX cents Is it when 5-10 people show up to your protests Is it when one guy goes on a hunger strike and gives up after a week because a doctor tells him "hunger is bad" These are not serious people and it's time for the media to start" [X Link](https://x.com/Dan_Jeffries1/status/1974596385082155042) [@Dan_Jeffries1](/creator/x/Dan_Jeffries1) 2025-10-04T22:03Z 32.6K followers, 64.5K engagements "Cheetah is crazy freaking fast. What the heck is this model After watching gpt-5-codex plod along for minutes and then this thing just lasers in and takes care of something faster than I can blink. Crazy town. Not sure I trust it for big work yet but man it can do rapid fixes like nobody's business" [X Link](https://x.com/Dan_Jeffries1/status/1978856085013291186) [@Dan_Jeffries1](/creator/x/Dan_Jeffries1) 2025-10-16T16:10Z 32.6K followers, 4063 engagements "It's not special cases. It's most of life. There are 10x more apps that have never been seen behind corporate walls that it would need to learn on its own. If the paradigm is that you have to train it on everything in existence it is doomed to fail. It's just not possible. Much better to build a universal learner that can learn on the fly continually. Like people" [X Link](https://x.com/Dan_Jeffries1/status/1979234122032640262) [@Dan_Jeffries1](/creator/x/Dan_Jeffries1) 2025-10-17T17:12Z 32.6K followers, XX engagements "Cheetah update after a day of working with it. Fast. Freaking so fast. Scary fast. I get the feeling the model is a diffusion model. It can read and write multiple files in parallel and spits out code and text like junior coder on Addy and five cups of coffee in time lapse. Seriously it was the first time I thought man when all AI is this f*ing fast how the f*ck do we keep up But it's sloppy. Super super super sloppy. You have to refactor every few steps if you try to do anything real. Awesome for lightning fast fixes. Awesome if you have lots and lots of code to reference. But doing" [X Link](https://x.com/Dan_Jeffries1/status/1979118309195813050) [@Dan_Jeffries1](/creator/x/Dan_Jeffries1) 2025-10-17T09:32Z 32.6K followers, 1838 engagements "So far it is very good at structured refactors with a strong plan and it absolutely rips through tasks like code commentary. I got pretty ambitious with it trying to craft/scaffold up a very complex langgraph/ts agent and it really really struggled. What I do is created super detailed plans with Claude/GPT-5/my own hand written instructions and then I have Codex analyze that doc and output a plan of steps. I literally just say: Read these docs at docs. Analyze next steps. Output a plan to make it a reality. I then sometimes just flip in the same thread to Cheetah to do it. Now it has the full" [X Link](https://x.com/Dan_Jeffries1/status/1979196696274882946) [@Dan_Jeffries1](/creator/x/Dan_Jeffries1) 2025-10-17T14:43Z 32.6K followers, XX engagements "All generally good advice. But different from mine. I never clear the context window until I have to and don't understand why so many folks think this is a magic solution to the LLM working better. To me it just makes it forgetful about the basics again. I run Claude's 1m token to the max as a supervisor/planner who doesn't code and context window lasts for days. I aggresively fill it up again with core docs and context in a fresh session start. In Cursor I rarely start a new thread unless I see it started to make stupid mistakes and then again fill it up with docs and plan outlines. Works" [X Link](https://x.com/Dan_Jeffries1/status/1976540199808037260) [@Dan_Jeffries1](/creator/x/Dan_Jeffries1) 2025-10-10T06:48Z 32.6K followers, 2121 engagements "This is what everyone in the industry keeps doing. They look at a job they don't understand and mistake it for a tiny subtask with no thinking. Even the call center example is not a great one. People have been trying to automate that forever. It's a series of a thousand little intelligent micro decisions and judgement calls on the part of the call center operator over the time horizon of the call which can be minutes or an hour. It also requires a ton of context gathering and interacting with proprietary systems" [X Link](https://x.com/Dan_Jeffries1/status/1979822659111211430) [@Dan_Jeffries1](/creator/x/Dan_Jeffries1) 2025-10-19T08:11Z 32.6K followers, XXX engagements "Brilliant podcast. 1) why are today's agents good at coding It's not text. It's high error tolerance and verifiable reward. It either works or it doesn't so you can incentivize that easily. If it messes up you just prompt again with no error penalty. That's why image gen coding and deep research took off. If it screws up you just prompt again. Can't do this in high error tolerance scenarios like surgery or driving a car or taxes. It's not text because good writing is not right or wrong. Can't reward multiple ways to do things. 2) Hinton picked on Radiologists because like many people in the" [X Link](https://x.com/Dan_Jeffries1/status/1979851090800558271) [@Dan_Jeffries1](/creator/x/Dan_Jeffries1) 2025-10-19T10:04Z 32.6K followers, 5808 engagements "I tend to look at jobs differently than most people so I think many jobs are far off though many tasks are not. Would you trust an agent now to gather up all the info for you taxes ask the right questions fill in the various forms and file it on one shot No. The only systems we have today are highly error prone which is why image gen deep research and code gen work. You can keep hitting the slot machine until it works. You can't do that with surgery driving taxes sending email to clients etc. I think most folks arrogantly assume many jobs are just brain dead but they're not thinking clearly" [X Link](https://x.com/Dan_Jeffries1/status/1979941422682849559) [@Dan_Jeffries1](/creator/x/Dan_Jeffries1) 2025-10-19T16:03Z 32.6K followers, XXX engagements
[GUEST ACCESS MODE: Data is scrambled or limited to provide examples. Make requests using your API key to unlock full data. Check https://lunarcrush.ai/auth for authentication information.]
@Dan_Jeffries1 Daniel Jeffries
Daniel Jeffries posts on X about faster, its just, build a, day of the most. They currently have XXXXXX followers and XXX posts still getting attention that total XXXXX engagements in the last XX hours.
Engagements: XXXXX #
- X Week XXXXXX -XX%
- X Month XXXXXXX -XX%
- X Months XXXXXXXXX -XX%
- X Year XXXXXXXXX -XX%
Mentions: X #
- X Year XXX -XXXX%
Followers: XXXXXX #
- X Week XXXXXX -XXXX%
- X Month XXXXXX +0.14%
- X Months XXXXXX +7.90%
- X Year XXXXXX +23%
CreatorRank: XXXXXXX #
Social Influence #
Social category influence social networks technology brands
Social topic influence faster, its just, build a, day of, a very, 1m, token, youtube, over the, all the
Top Social Posts #
Top posts by engagements in the last XX hours
"At what point does the cognitive dissonance kick and you realize your AI Doomer movement is a joke Is it when the 'If You Build It Everyone Dies' falls off the best seller list in a week and you realize your astroturfing foundations that bought enough copies to put it there for that one week are the only copies getting sold before it ends up at Goodwill for XX cents Is it when 5-10 people show up to your protests Is it when one guy goes on a hunger strike and gives up after a week because a doctor tells him "hunger is bad" These are not serious people and it's time for the media to start"
X Link @Dan_Jeffries1 2025-10-04T22:03Z 32.6K followers, 64.5K engagements
"Cheetah is crazy freaking fast. What the heck is this model After watching gpt-5-codex plod along for minutes and then this thing just lasers in and takes care of something faster than I can blink. Crazy town. Not sure I trust it for big work yet but man it can do rapid fixes like nobody's business"
X Link @Dan_Jeffries1 2025-10-16T16:10Z 32.6K followers, 4063 engagements
"It's not special cases. It's most of life. There are 10x more apps that have never been seen behind corporate walls that it would need to learn on its own. If the paradigm is that you have to train it on everything in existence it is doomed to fail. It's just not possible. Much better to build a universal learner that can learn on the fly continually. Like people"
X Link @Dan_Jeffries1 2025-10-17T17:12Z 32.6K followers, XX engagements
"Cheetah update after a day of working with it. Fast. Freaking so fast. Scary fast. I get the feeling the model is a diffusion model. It can read and write multiple files in parallel and spits out code and text like junior coder on Addy and five cups of coffee in time lapse. Seriously it was the first time I thought man when all AI is this fing fast how the fck do we keep up But it's sloppy. Super super super sloppy. You have to refactor every few steps if you try to do anything real. Awesome for lightning fast fixes. Awesome if you have lots and lots of code to reference. But doing"
X Link @Dan_Jeffries1 2025-10-17T09:32Z 32.6K followers, 1838 engagements
"So far it is very good at structured refactors with a strong plan and it absolutely rips through tasks like code commentary. I got pretty ambitious with it trying to craft/scaffold up a very complex langgraph/ts agent and it really really struggled. What I do is created super detailed plans with Claude/GPT-5/my own hand written instructions and then I have Codex analyze that doc and output a plan of steps. I literally just say: Read these docs at docs. Analyze next steps. Output a plan to make it a reality. I then sometimes just flip in the same thread to Cheetah to do it. Now it has the full"
X Link @Dan_Jeffries1 2025-10-17T14:43Z 32.6K followers, XX engagements
"All generally good advice. But different from mine. I never clear the context window until I have to and don't understand why so many folks think this is a magic solution to the LLM working better. To me it just makes it forgetful about the basics again. I run Claude's 1m token to the max as a supervisor/planner who doesn't code and context window lasts for days. I aggresively fill it up again with core docs and context in a fresh session start. In Cursor I rarely start a new thread unless I see it started to make stupid mistakes and then again fill it up with docs and plan outlines. Works"
X Link @Dan_Jeffries1 2025-10-10T06:48Z 32.6K followers, 2121 engagements
"This is what everyone in the industry keeps doing. They look at a job they don't understand and mistake it for a tiny subtask with no thinking. Even the call center example is not a great one. People have been trying to automate that forever. It's a series of a thousand little intelligent micro decisions and judgement calls on the part of the call center operator over the time horizon of the call which can be minutes or an hour. It also requires a ton of context gathering and interacting with proprietary systems"
X Link @Dan_Jeffries1 2025-10-19T08:11Z 32.6K followers, XXX engagements
"Brilliant podcast. 1) why are today's agents good at coding It's not text. It's high error tolerance and verifiable reward. It either works or it doesn't so you can incentivize that easily. If it messes up you just prompt again with no error penalty. That's why image gen coding and deep research took off. If it screws up you just prompt again. Can't do this in high error tolerance scenarios like surgery or driving a car or taxes. It's not text because good writing is not right or wrong. Can't reward multiple ways to do things. 2) Hinton picked on Radiologists because like many people in the"
X Link @Dan_Jeffries1 2025-10-19T10:04Z 32.6K followers, 5808 engagements
"I tend to look at jobs differently than most people so I think many jobs are far off though many tasks are not. Would you trust an agent now to gather up all the info for you taxes ask the right questions fill in the various forms and file it on one shot No. The only systems we have today are highly error prone which is why image gen deep research and code gen work. You can keep hitting the slot machine until it works. You can't do that with surgery driving taxes sending email to clients etc. I think most folks arrogantly assume many jobs are just brain dead but they're not thinking clearly"
X Link @Dan_Jeffries1 2025-10-19T16:03Z 32.6K followers, XXX engagements