AI Coding is here... What’s next? My 2025 GenAI BIG BETS (AI Agents, o3, Paic)

[@travis1687]

Just wanted to say you've changed the way I think about all of this and it's helped me a ton. Thanks.

[@Djolesabijacica]

Wow, what a video! As a mid-level engineer who uses AI daily, I can confidently say that I started outperforming the senior members of my team long ago. They avoid using AI, believing they don't need it because they can handle everything themselves. I began experimenting with AI tools like Bolt and Lovable, and in just a few days, I managed to create a full-stack website that is now live-something that would have taken me months to do manually.

[@ryantigi2583]

I would like to ask, what is the perfect fit for using AI for coding I tend to overuse it especially since most of the models launched just around when I was getting started to coding myself.

[@2034-SWE]

@@ryantigi2583no such thing 10 years from now, 99.9999% of humans won’t be writing a single line of code by hand. Not even a character The sooner you figure out how to do this with bug-free accuracy, the further ahead you’ll be of almost everyone on this planet rn

[@BrianMosleyUK]

Can you share the address of your full stack website?

[@indydevdan]

This is the curse of the experts mind. Just like you mentioned "...they don't need it because they can handle everything themselves". This is the fastest way to self deprecate in tech and burn out a great senior+ advantage. Great judgement from you deciding to leverage GenAI. This story only ends one way.

[@Mkp-v3o]

Just finished the 3. course and I want to thank you for creating this value Dan! I am developing with AI for a long time but your prompts and workflow helped me to gain different perspective and cant wait discover more! as others have mentioned, it would be nice to have a community of like minded people.Ones again thank you and happy new year for every one!

[@jamespistell2689]

Just bought your course. Even as a web dev with little Python experience I still found it easy to follow along. The spec prompts really have made me change the way I think about writing code. I think I need a moment for this to settle in my brain. Thanks Dan and happy new years.

[@indydevdan]

Happy New Years! Spec prompts are insanely powerful. This is how you scale up your AI Coding Assistant.

[@JimmyJones-mo2to]

Dan is definitely NOT a grifter. I went through the course and finished up yesterday, it was TOTALLY worth the price. We better start figuring out how to use the tools, let go of the how, and start focusing on the what, or we are going to get left behind. Get the course!

[@JakexGaming]

I know nothing about coding but love learning about AI and what’s to come. Thanks for putting this content out!

[@kora5]

I bought the course based on the value of content you already provided in all your free videos. Thanks for all your insights here and before.🙏

[@benjaminadams735]

bro why did you have to drop this absolute heat on us like that? Absolutely awesome, thank you for all the helpful information, poetic in explanations and teaching to all. Your work is class, thank you Dan

[@RuneX_ai]

Thank you for the course.

[@sanjaygupta4]

Best video for the year end. Thankyou brother.

[@TreeLuvBurdpu]

"Thinking in Bets" is a great book by Annie Duke. Are there any more game-theoretical or code oriented versions?

[@indydevdan]

Thinking in bets is one of my favorites. Highly recommended. Everything we do is a bet. Every decision, every action, every second spent on THIS instead of THAT is a vote/bet for something and against something else.

[@AnimusOG]

I needed this Dan, you are my hero!

[@ModernCentrist]

This video is eye-opening. Thanks for the great content.

[@JustinLietz]

i started watching you a couple months ago, and I've been almost obsessively learning as much as I can about AI and agentic engineering workflows, putting together ideas and strategies to benefit from this wave of new technology. I'm trusting you with your course, just bought in now

[@93cutty]

Dan literally makes Mondays worth getting out of bed for.

[@93cutty]

And also, do you have a video on 'maxing out your compute?' I've been inactive and just watching your videos but since you first started talking about agents, I've been wanting to do that. I have a lot of 'starters' or like an openAI Assistant to use, or running n8n, etc. I know 100% it's definitely the most important thing. I'm tired of sitting here and tinkering.

[@indydevdan]

Maxing compute is a big theme for the channel in 2025. Stay tuned.

[@gerryscat]

I think hope eventually we have like a drawing design tool for creating UI, APIs, inputs and outputs (schemas). Probably need to add some details, like volumes, user roles, things I haven't though of yet, etc. Then AI deploys it.

[@Daevalus]

"there is no wall". We can already see this taking form with Deepseek and MoE.

[@ryantigi2583]

That was definitely a shocker especially after their release of their R1 Lite Preview reasoning model

[@philq01]

I just signed up for Principled AI Coding

[@danielmayo2]

I've been going through your course, and it's cool to see me using Aider exactly like you're doing, but the cost being a fraction of what you're showing because the reasoning model costs have just gone down so much.

[@kilianlindberg]

Much appreciated, keep it coming 🙏

[@superdrew907]

Am I the only one that had to pause and rewind? I wasnt gonna sign up until it was confirmed that I was absolutely gonna Learn HOW 2 SUCK, this is gonna be great something I've been putting off awhile now, I'M aLL IN!

[@FalconStudioWin]

21:00 i think that it is not data that will be leading in better models as we have smarter ai systems but rather user experience the whole system brings and how it is used for user to be most effective ever

[@mistn]

Just purchased the course!

[@dreamineering]

Every engineer needs to be able to think like an investor and every investor needs to be able to think like an engineer. Both need to place bets to position for maximum leverage while avoiding unserviceable debt AI agents can think 24 7 fully plugged into the matrix of data to react to updates to act on valuable insights

[@RuneX_ai]

Thank you dan for these insights, very very important. Will remain critical your predictions, but still, i feel like walking around with a filled bag instead of no clue

[@mindcastsoftware]

I wonder if the Evergreen AI Coding Experience described in the video at 10:12 could be a codebase that comes with its own AI. For instance, if something like WordPress came with its own AI, then you could just tell it to create plugins for itself.

You are both smart, good and generous! thank you..

[@sjkba]

I love your content and that you're putting yourself out there. My low confidence, high impact bet would be that XAi will come out with a model that really crushes openai for a while. The amount of compute Elon has is immense.

[@jksoftware1]

One thing that hinders our access to these new LLMs is OpenAI's very bad policy of how it allows API access to the most recent models. For example, API access to o1 has been a dumpster fire. Not even all of tier 5 has access to it let along tiers 1-4. o1 Pro mode is ChatGPT only which can be a pain in the butt too.

[@__5545]

????

[@Samuelkings]

I am coding again thanks to AI, Just got Cline in my VScode and got Deepseek API, also got Ollama downloaded.

[@champceo1581]

Hi Dan, I've purchased your course and currently in the midst of completing the 2nd lesson. Is there any channel that I can ask questions when I have issues?

[@indydevdan]

Definitely. On the course page you can click "Ask a question". If that doesn't work (no email enabled on your os) scroll to the bottom and copy the email manually and shoot me a message.

[@cariyaputta]

True. The general public vastly underestimate AI researchers, they rarely read the actual bleeding edge papers and just jump on the hype train when it's there. There's a ton of papers pumped out recently about why RAG will be obsoleted due to the emergence of giga high-context LLM/LCMs. The frontier of AI researching is pushing really hard without much knowledge from the public. For example, the research on optimizing KV Cached with Infini/Ring Attention and Inference Time Compute.

[@hypexeyy3651]

I’m a 21 yo student with a finance background, diving into the AI world over the past few months as I see its potential. Is this course worth it for someone like me, or should I gain more experience to truly maximize its value?

[@bzzt88]

Just got the course, mostly to keep up. I haven’t coded in years, but why shouldn’t I build my own stuff at this point?

[@DaviiRolim]

I love your content. It would be awesome if we entered a community when buying the course. It would be nice to exchange ideas with other like minded folks.

[@aaronabuusama]

on a level the course 🔥🔥🔥🔥🔥 the lessons up to the last two were cool, nice hints and tips and all in all worth the money up to that point, but the last two lessons!!!!!!! my only gripe is there is no discord or tg

[@travis1687]

biggest AI prediction for me is xAI, based on the fact they just built the worlds largest super cluster which they want to continue to expand. Elon already has government contracts and a great relationship with the president elect. Regardless of how you feel about politics, everyone is sleeping on xAI.

[@augmentos]

Love it good accountability. AI code this year gonna change the game. I’m ONLY hiring ai devs that are dev based

[@Nairb932]

Is this course just for Programmers / Code Engineers or is it no code brainlet friendly ?

[@bradleyfraser4026]

best video yet

[@blackwhite-3607]

oh mans didnt realise the price increase was today... :( purchased nonetheless. looking forward

[@biochemcompsci]

The $COIN was good vs SOL and ETH. Interested in the subscription

[@bjkang70]

Does the 'master ai coding' course include the 'vision grid' coding course?

[@j.hanleysmith8333]

Cursor team is so f****** cracked and their software is on a crazy improvement clip. Good ROI to just use it move on

[@mattmiller4978]

Agreed they are cooking. I am all in on Cursor after seeing them on Lex, then acquiring Supermaven

[@LukePryor]

AI Agents as a Service could definitely happen and is probably not far off - very good idea

[@indydevdan]

Honestly this idea is so cracked. Big potential.

[@augmentos]

Good vid and predictions. I was disappointed to hear you call Llmama Open Source, it is NOT! It is open weights! Let’s stop giving a trillion dollar company a free pass and bestowing Open Source on them when they ARE NOT

[@iliyaisd]

I think it's debatable whether OpenAI is in the lead honestly. Claude Sonnet either on par or beats it in many practical use cases. Even despite o1/o3. This is just IMO.

[@noirblanc9446]

I kind of did this context building!

[@App-Generator-PRO]

I'm doing all of this today. I know people take poe not seriously, but as soon as we smashing that wall people will. And everyone can do this. Even I as a stupid "no coder".

[@bambo999]

2025 is going to be a blast, if you are into whole AI world 😎

[@ManWhoLostTooth]

Maybe in the future we can all have our own courses to sell.

[@paraconscious790]

Unfortunately $150 is still considerably high in India. Waiting for India specific pricing.

[@loryo80]

In 2025 Gemini will take the lead of mass developer adoption, video generation will be democratized, Xai will bring new perspectives with Tesla history in autodriving and the massiv capacity of calculation built by xai Openai with O3 will build the foundation of intelligence access class, so entities with large financials capacities will win the game, We will assist to the extension of the geopolitic to the ai access

[@PazLeBon]

Same as anything for sure, big dogs eat cats

[@MilosKvakic]

I recommend everyone to find the book titled The Hidden Path to Manifesting Financial Power, It changed my life.

[@stevecaya]

Prediction: developers are wasting time with small start up AI company tools. The big giants like Microsoft Foundry will eat them all. My bet, learn foundry.

[@TreeLuvBurdpu]

That's kind of dark. Microsoft always makes the broad but poorly composable tools. Do you think Foundry is different?

[@stevecaya]

@ to actually deploy these apps you need the infrastructure. The business world is not going to trust some cool one off app, they want the security of Microsoft who can deploy at scale. GitHub just added the AI coder for free to keep people on their ecosystem. And don’t be fooled, open AI is Microsoft.

[@indydevdan]

Dark but I kind of like your prediction. Problem is big tech will always be slower and less innovative than a fast small lean tech company. It's the nature of large systems. Incumbents have built in risk to manage.

[@peaduaw1]

walled gardens coming back big time

[@larsfaye292]

Front-end roles did absolutely not disappear. Come on man, there's nothing to support that and you blew right by it with nothing to show for it. Every single front-end dev I know is not only employed, but busier than ever!

[@ryantigi2583]

As a university student focusing on AI, I understand how crucial it is to specialize. However, I keep asking myself: How do I decide which field to pursue? Several high-potential areas have caught my attention-mentioned by AI researchers and CEOs-including digital biology (e.g., Neuralink and related applications), the space travel industry, and advanced AI research, particularly AGI and ASI at higher levels (levels 4 and 5 on OpenAI’s scale). Each field is fascinating, and it’s intimidating to commit to just one, especially when they all hold such promise. The challenge of choosing feels both exciting and daunting.

[@jayazhane7374]

you could try what I accidentally did and specialize in something completely different. so I specialize in Japanese which has slowed me to differentiate myself in the tech industry. I can learn new tools and branch into any tech industry I want, because I can position myself as the expert on doing that thing IN JAPANESE which has been invaluable for companies.

[@ryantigi2583]

​@@jayazhane7374 wow that's actually an amazing, but unexpected approach, I will definitely keep this in mind and have it sort of as a goal or target. Thanks for the Insight.

[@indydevdan]

My advice: Pick two, go absolutely hams. Then slice down to one after a year. Spend most of your day focusing on these two things and I mean literally count the hours and if it's not more than 50% of your day scale it up. Make sure you're investing in yourself. At some point it will become clear which one you like more and which one has a better career path for you. Make the hard decision. Then commit. The biggest mistake I see noobs make is 1. 'trying to do it all' 2. 'specializing too late' 3. 'fear of commitment'. Whenever someone says 'you're young you have time'. Ignore them. Get skills now. Time is your greatest advantage and your greatest enemy.

[@2034-SWE]

👋

[@MrSarkoko]

Merci pour cette vidéo, et surtout pour l'avoir mise en français.

[@indydevdan]

Avec plaisir. Merci de me faire confiance. Cette technologie doit être diffusée au-delà des frontières et des limites.

[@lel7531]

Making it $150 is literally insane and also saying it will increase is preying on people. Not many have $150 laying around to spend on a course like that. What is the percentage of truly Original content and ideas compared to content that has already been posted ?

[@Codemanlex]

Personaly I do not have that much to spend The truth is you dont need it atleast for now. Free resources online Learn on your own If you got cash later on you can then buy a course

[@2034-SWE]

$150 for 2+ years worth of experience distilled into

[@remsee1608]

Open source has surpassed closed source and in 2025 should see the beginning of automated improvement, aka the intelligence explosion, aka the singularity

[@PazLeBon]

Lmao

[@remsee1608]

@@PazLeBon what’s funny

[@alexfrink6694]

a bold prediction would be open ai totally blowing up b/c they can't keep up w/ the spending others are making. yes, they've got msft as backers but eventually investors will balk at the spending w/o roi. 03 is outrageously expensive and maybe only temporary when other ai models catch up. sonnet 3.5 is better than 01 in my coding experience and it's free now. the only profitable ai company seems to be tsla. i think you might be totally missing the boat on grok 3. we should know how good 100,000 gpus are when it comes out in jan/feb 2025. if fsd 13 got that much better being trained on more gpus is 1 data point and grok 3 will be another data point to see if scaling laws exist. i'm a 1st time watcher & a msft dev. i'll check out some of your other videos b/c i think you're spot on in a bunch of places.

[@remsee1608]

OpenAI simply has fallen behind. GPT-4 is inferior to Deepseek and Claude, Sora is inferior to Google Veo. Additionally, we will see xAI emerge with Grok 3 in 2025

[@naranbaz]

Elon overpromises and underdelivers.

[@remsee1608]

@@naranbaz Other than full self driving, he has delivered on everything he has promised, albeit not always on his aggressive time horizons.

[@naranbaz]

@@remsee1608 Im rooting for anthropic, as i like that model the best.

[@andrecummings818]

I think you’re sleeping on grok when saying open ai will stay king. Elon being the leader or grok idk cause he did make a 100k gpu data center in 19 days which was considered impossible to do and still hasn’t been replicated yet. And he’s planning on going to 1m gpus. If the scaling laws hold up he’s booming past everyone even if they don’t he still has way more compute and the engineers he has is probably on par with open ai’s. If he needs better engineers/researches I doubt he can’t find talent, it’s Elon

[@indydevdan]

I'm eyeing grok but I don't see any real innovation yet on the LLM side. Just insanely impressive data centers. I need model stats.

[@andrecummings818]

Fourth

[@MugiwaraNoDeji]

Google will surpass OpenAI

[@victorbjorklund]

in what way? Users? State of art models?

[@MugiwaraNoDeji]

​@@victorbjorklund SOTA models, they pioneered the market with Transformers, have the most data(both in quality and creativity in data), and have a built in ecosystem already. But, the main reason I say they will win is because they are focusing on games. This is the best way to train models. Games simulate systems and are unlimited in the way information is structured(basically a infinite synthetic data). A game represents basis of life and our universe, a system that is created based on rules that is fractal in nature. Google is developing a model to create games and another model to play any game(basically mastering novelty). They already have alphaGo, and our SOTA models are good at skill acquisition and mastery. Imagine a model that can play any game and become a master or pro at any game? That is us, a human, or more specifically every human. The model that can create games will help expand the scope, simulating harder systems which corresponds to solving the hard problems in our world like fusion, quantum computing, cancer, etc... Its all about systems, life is a game, the universe is a massive game. I think the issue we have is we think a game is just a game, and dont realize how games truly correspond to the nature of our universe.

[@indydevdan]

Google is definitely waking up - I can see it but I'm willing to bet against it. Why? OpenAI keeps leading the market. Everyone else plays copy, catch up and improve (which is a great strategy don't get me wrong). It is MUCH easier to copy and improve (cough cough apple) than to innovate back to back to back like OpenAI has. Let's see how this game plays out.

[@Kevencebazile]

Second

[@pro-coded]

First

[@JohnLewis-old]

@indydevdan Why no love for Cline?

[@janlazorik]

Third

[@lel7531]

Making it $150 is literally insane and also saying it will increase is preying on people. Not many have $150 laying around to spend on a course like that. What is the percentage of truly Original content and ideas compared to content that has already been posted ?