Podcast··42m

Episode 05 — OpenAI Just Raised $122B — AI Is Getting Out of Control

OpenAI's record $122B raise at $852B valuation, agentic AI momentum, humanoid robots entering production, and why the next AI products will be autonomous.

Listen onYouTube

Episode notes

OpenAI's record $122B raise at $852B valuation, agentic AI momentum, humanoid robots entering production, and why the next AI products will be autonomous.

Chapters

  • 1:00 — Overview of OpenAI's massive funding round
  • 5:00 — Discussion on AI-driven business run by one person
  • 10:00 — Exploration of AI policy and potential shifts in taxation
  • 15:00 — Analysis of humanoid robotics advancements

Transcript

Note: this transcript is auto-generated and lightly edited for readability.

Tim: welcome welcome to another episode digitalized lab by digitalized agency actually episode number five time is going by real quick we got lots to talk about again this time we got an online episode next time we will be back at the usual location but we hope you guys enjoy it and still like it give us feedback and let's jump right into it and by the way the weather is great

Sebastian: just wanted to say even even in both locations i mean the reason why we are recording online is because we're not all in in harlem so but i see in both locations really nice weather beautiful

Finn: indeed indeed yeah it's it's a rare occasion but uh we'll take it as it is but yeah let's jump right in um i'll just take the word because i have uh quite a quite an interesting topic um

Sebastian: no i mean a lot of things happen again right it's crazy how it moves so fast every week we say it happens a lot but it actually does and i think we don't even cover 10 percent probably not even 10 percent yeah that was way too long but sorry yeah continue finn

Finn: no i just i just wanted to start with the the open ai uh private funding round so the biggest one ever on this planet earth they raised let me check the numbers 122 billion dollars which is a crazy amount uh just raised it at a valuation of 852 billion so quite crazy whNoat i think is quite interesting is even amazon pitched in the biggest amount actually of 50 billion uh which is interesting because they were the first one of the first investors in entrophic yeah and now they kind of playing both sides yeah yeah yeah definitely but now they're they're kind of pitching both sides and at the same time they still don't really do their own ai they just buy in so yeah interesting

Tim: yeah i think it's so interesting to see it's so interesting to see that ai momentum does not slow

Sebastian: down everyone was still scared wait wait wait can we can we first start with with the freaking evaluation that is by far the highest evaluation that is not even at the stock market they didn't even have their ipo imagine what was going on when when they launched their ipo no they're getting

Tim: it i mean they're not they're not far from profitable they're burning money uh on their hourly basis yeah millions so yeah it is it is quite quite interesting to to see definitely yeah yeah i know

Finn: it's also the first time ever that retail investors got access even though it's a private uh funding ground so that's why i guess i i think they just want to push their ipo a bit before they actually go go public uh but because they also need their money they they they're burning through cash as tim said already uh heavily and yeah they're making a big losses right now but yeah they're getting

Sebastian: funded so for now they're good but yeah they're getting the money to burn it so why why why why not

Finn: fair enough enough but yeah no i just uh thought that was quite interesting and they're actually planning to go public uh around q4 2026 so q4 yeah let's see let's see who is first

Sebastian: and traffic or open ai did you actually you're gonna be interested i didn't hear anything if they if they want to go well they're they're yeah they definitely need to or they want to because there's some different level of money i would say um but yeah both both said um they want to ipo 2026 and i mean 2026 if it's actually if actually all of the companies say that said that they're ipoing then it's going to be a crazy year spacex or yeah it's also one of them

Tim: we also got uh spaceax i think yeah end of this month actually they're they have the apos yes yeah i'm pretty sure i don't have the exact numbers in mind but i'm pretty sure they they said end of april quickly check they already published the date uh yeah check check for us um so yeah looking forward to that as well but uh yeah let's see let's see but it's also really

Sebastian: interesting like you said earlier um finn that that microsoft it was microsoft right who is invested in both um amazon thank you um amazon is is invested in both in open ai and in entropic that i find interesting yeah they don't really put anything on oh for sure they they put the basket not even not only in one no they put don't put the eggs in more than one basket yeah that is the same

Finn: but i think it's interesting in how much money they actually raised because what what was it 122 billion at a valuation of 850 something so if you look at i just got the numbers for for spacex they're targeting a valuation of 1.75 trillion which is completely out of this world but even with their valuation they aiming to to raise about 75 billion so less than open ai with 800 which is not even half of the valuation interesting well we'll see how it is but uh yeah i mean it obviously they're trying to merge with xai and everything so that's a different story but uh as tim was actually nearly right uh they're planning to go for early june so june 11 early june okay my bad uh an event yeah i mean close to know close i get to know gotta put that in the cover let's just uh not to be honest i can expand a bit on the topic sorry go ahead yeah i just wanted to say now that we

Tim: missed uh the next ipo like with figma yeah but i'm glad that i did not did not have yeah but the first

Sebastian: if you would have got some some shares at the first bull run then why not sure would have left it would have needed to be nice but probably not so better yeah no go ahead tim actually that brings

Tim: me to my to my second second topic because we're already talking about the big numbers i remember that conversation do you remember remember around one year ago we talked about that there will or actually sam altman did he said oh there will be a company run by one person which is uh one worth one billion and we're actually there actually a week ago the first person got published yes his name is med medevi or something like that no matthew matthew callahan and yeah he spent roughly uh twenty thousand dollars on api cost vibe coded something together and since running his business with um i'm not sure how many employees i think like three if i'm not wrong three employees and they already reached a revenue last year of four uh 401 million uh us dollars and are set to try to breach the the the the holy number of 1.8 billion in sales in 2026 damn crazy what are they doing what what is the same question um what's the secret sauce is they have a couple um a couple ai tools which they um which they released and it's also like a mixture out of code content ads support and internal analysis

Finn: systems for basically a whole ai suite yeah yeah crazy okay do you know what the comment oh okay so it's

Tim: actually just says his uh brother elliot that's the only only employee yeah wow and they spent uh about

Finn: twenty thousand us dollars i mean that perfectly shows that it's definitely possible to to run a

Tim: two-man show nowadays making a shitload of money yeah so it's definitely the cheap ai tools fast build

Finn: tiny headcount and that translates into future revenue age yeah but isn't it funny also on the

Sebastian: other hand it's almost like he would have paid someone to build him the software for the 20k you know what i mean the only reason why he probably could have achieved that is because he could iterate on its own so or he could iterate on the software and probably talk with users i guess i don't know haven't looked into it and through that it made it way better than just paying 20k to to someone else and that shows again the hyper personalization so for real that ai helped with just iteration and

Tim: making people faster and not replacing yeah and that's that's exactly what i also put down in my notes is kind of like the old model was to hire people first then of course scale the software and so on and now it's kind of like you skip to people you just build and scale

Sebastian: yeah let's go sure sure yeah

Finn: interesting yeah really good let me just do you guys go back to to the open ai topic for one second because they actually dropped some other interesting stuff um yeah different like not not ipo but um basically they wanna they have they published like a yeah industrial kind of policy paper for the intelligence age no yeah it's interesting because the three kind of main points from it are firstly they want to tax robots which means they want to tax companies that replace people with robots in the same way as if they would hire people so in that sense they want to kind of solve the problem with who get like how do people get paid if the jobs get laid off so they basically want to tax them and then build a public wealth fund which is yeah kind of hosted by the biggest ai companies also kind of sus and uh they would then yeah they would then redistribute to every citizen so yeah kind of interesting and they're pushing to get people not even companies governments to push for the four day week at full pay to kind of test it i think i think we're not

Tim: we're not that far anymore of uh where we will stop um taxing taxing uh work in in the way we currently do i think we're gonna see a shift from taxing humans to actually only taxing ai that i'm pretty sure we're not we're not that far anymore we i don't think we're gonna tax labor like physical and human labor um that long anymore and i think this is gonna get be there much faster

Finn: than we think but it will be it will be interesting times for sure i don't know how long it's gonna take of course but the speeds that things are going right now at least in america it's gonna gonna

Sebastian: become a test case for sure and i also think that is tax for real the the next step you know what i mean why use the old system no i guess crazy but i mean with that big of a shift then why use tax overall and just redistribute wealth on its own because if we actually kind of get the abandons of compute and electricity and all of that stuff that we actually need for ai then ai can just do everything and humans don't actually need to work anymore and no one needs to work anymore because people don't need to pay for the bus because the bus gets operated by an ai and the ai gets operated by basically electricity and we have electricity on compute or mass so no one has to do anything

Finn: anymore but then you don't need taxes anymore yeah i guess taxes are just a word that people understand you know because you yeah true you need to you need to have some i don't think the world is going to end up with no currency overall i think there needs to be some form of exchange of things but yeah it is an interesting topic if i mean not not every job is gonna gonna fall apart definitely not it's just those repetitive kind of jobs those factory kind of jobs those are actually kind of impacted by robots in a way for sure they already have been to be honest if you look at those big industrial robots if you look at bmw factories or something it's nearly all of them like tesla for sure nearly all of the workers are robots but then there's still people operating those robots blah blah at some point that's gonna change and be i don't know fully ai but i can't uh can't tell you what

Sebastian: it's gonna be and or i think they're still doing that in a test factory basically and they partnered up exactly we had a presentation about that too they partnered up with figure ai also a really nice humanoid robotics startup from america and they are also building basically the next humanoid robot really nice and they partnered up with bmw to operate one station of their car plant and 24 7 because they basically had i think six robots in total and always a shift of three robots that means as soon as the battery went low they basically changed the robots and then they locked themselves in so they could charge again and then after i think six hour shifts or something they basically swapped out and the error margin was 98% less and i think some other really impressive notes but it's just too expensive for right now i mean those numbers could be the answer for right now yeah yeah but it is fun because it is basically robots using other robots then again or not again but this is the next step

Finn: was it figure that have those the robot that kind of just has this battery and he just takes another battery off the wall and just changes it like himself i saw that i saw that as well that they don't even switch around but they literally go to a wall with like batteries and just swap the battery and then they just like their own to to charge again and they don't even have to switch i don't know if it was figure but maybe it was a different no i saw this thinking which one it was

Tim: not sure i have to check on it uh whatever what's your what's your top five uh figure figure figure no figure sorry humanoid robot companies

Sebastian: if you had to list them shoot them all the top five is crazy top three i think i think it is actually figure and sunday robotics i think they those two are on on place one or two i don't know but they're both nice basically doing the same but sunday robotics makes it their robot looks like a lego figure figure

Finn: and then it's just awesome what is the chinese company called again they're also having like

Sebastian: unitry yeah those are also quite interesting they're also really one the two one yeah yeah and especially they're really good um but but there's also and now an open source humanoid robot by someone i forgot his name but it's super nice yeah he he basically sells the kits for 50k and everything to do yeah yeah super nice and you can follow his whole progression and everything is open source you can basically just buy the kit and build it yourself and do the development on it and it's actually if you if you think about it it's 50k it is relatively cheap because we're still right at the beginning and imagine that i would love to buy that crazy yeah no yeah no it will be interesting but i would say figure unitry and sunday robotics and or if you if you want to say boston dynamics is a company than that but they're more research

Tim: lab i was about to say it as well um i think boston dynamics

Sebastian: yeah boston dynamics is just too good and brings especially the actual stuff like that is i went down the rabbit hole once about actuators and it's crazy it has opened a whole new world but yeah yeah i

Tim: think also the the movement of the of the boston dynamic ones i'm not sure what the name of the specific robot is um i have to look it up but i think the move no no the humanoid robot yeah the humanoid robot one that does like flips and stuff yeah with the disc face that one yeah i think the

Sebastian: movement is quite insane yeah but actually the chinese models are are almost there you can see that i saw a really nice reel which compared the humanoid robots because china did some i don't know some some showcasing or whatever and had i think 50 humanoid robots dancing and they did that last year so 2025 and 2026 and the jump from the level of how they're moving it was actually crazy i hope i can find that later and then i will put it into the description but it was actually amazing to see they the first video was actually like you imagine how a robot would dance and the typical robot move and then the second video was actually like the boston dynamics they jumped around they did the backflip

Finn: and it it looked really human it was kind of scary you guys see that boston dynamics got acquired a while ago boston yeah by who they got acquired by who hyundai interesting obviously crazy try to implement it in their factories for 1.1 billion by the way and they actually found in 19 1992. but i mean they only

Tim: call the i mean they didn't did it like a full acquisition now it looks like it for one billion like it i would have bought it for a billion that seems true but you just want to say

Finn: a quiet robotics company and 1.1 billion deal okay i don't know i didn't really look into that but uh interesting to be honest yeah but but one thing

Sebastian: that i forgot earlier about open ai they they also made a big deal with the tbpm the the podcast show did you guys see that oh my big girl miles

Tim: oh my big girl miles i've actually never watched an episode of them no same not there that's i do watch them sometimes but what's interesting with them is they're not a uh that traditional podcast necessarily they're much more right yeah they're they're live streaming and then they're cutting their clips into little podcast episodes and of course they feature um alex carp the CEO of palantir all the all the big names quite often yes and as we just said they got acquired by uh open ai they did not disclose any specific numbers but it was just said that they will have help open eye with marketing and communications and i think that is definitely a right move so if any of the of the bigger companies out there are also looking for a podcast to take over

Sebastian: nah i know i don't know

Tim: no of course not but i got i yeah yeah i have something i want to talk about go ahead i read it i read a um a research article by gardner the analytics company and i think it was really really interesting so what they basically said in that so it was about um ai agents and um their main forecast was pretty much that by 2028 more than half of all enterprises will stop paying for assistive ai tools so co-pilots everything completely scrapped by half of all the enterprises but like and i think that is half of them yes completely stopped they will all move to agentic platforms so so does that mean that normal

Finn: human employees won't have access to any ai anymore or will they then work with well i'm guessing they

Tim: no of course of course they will but they will move to agentic platforms so and not to just you know is there a definition of agentic um i would need to look it up again but uh how they defined it but i mean the general uh definition of agents right is you have a you have a start and a goal and the agent figures out themselves how to get there rather than and can like call tools and so on rather than just um um dull copilot just give you an yes that just pulls up an information or so yeah but some of the key key numbers that i that are uh rolled out because it was quite interesting so in 2025 only five percent of enterprises enterprise apps had ai agents only five percent last year and they predict that by the end of this year it will jump jump eightfold to 40 percent already crazy 40 percent from five to 40 percent and at least 15 percent of day-to-day work decisions uh will be made by agentic ai by 2020 15 15 sorry 15 okay yeah that makes more sense but i think that is in two years in your two years 15 of day-to-day work

Finn: decisions are made but is that based on america or is that based on because worldwide i think that's a bit at least from what i see is a bit far but if it's only america kind of later then it would make sense

Tim: let me check in a second let me check in a second yeah and i will come back to you

Sebastian: but it's still impressive the the eightfold for me yes i mean from five percent to i mean if you just imagine how much more compute that is that is also an eightfold i mean it's switching around but of course

Finn: agents need quite a bit more than uh yeah especially with context when they're on no the prices won't get better any soon with ram no no just by ram i guess no but i think it's not even ram i mean you

Sebastian: also need to store all of that so hard drives i mean they are also skyrocketing electricity don't even talk about it but it's actually i found a really fun post by a tiny grad do you do you guys know that i think i already mentioned it but super fun small company and they're basically focus on ai boxes i would call them computers specialized for ai and they have some some smaller um power houses with i think one is four times rtx 6000 and stuff like that so actually heavy work with a lot of we uh vram to run extra fun llms on it and and i saw earlier they are now building um a whole container full with compute power that you basically can order and you can now play uh put yourself into the waitlist for 100k and the container should go up to 10 million imagine that and you can just basically deploy it everywhere you just need one block yeah yeah yeah the mini data center yeah exactly yeah mini data center i just wanted to say but still imagine how cool that is you can almost everywhere plug that i mean maybe not with a 10 million budget but super interesting i mean you have private kind of

Finn: processing and stuff it's an interesting yeah i think actually that is the future the especially for

Sebastian: sms or whatnot to to get some kind of off-grid compute so that is actually yours and that is then basically what what makes you unique or your unique salary point your agent behind it i mean jensen who said it every every company needs a genetic change management now change management was my course yeah

Finn: yeah when i need to have a look at my heart yeah oh interesting but perfectly leads over to did you guys hear of project conway no actually not any of it centropic they're leaking new stuff again as every week new stuff so they were yeah they were surprisingly quiet yeah sorry go on yeah well i mean the last last episode we talked about how they unfortunately leaked uh the whole source code of cloud code so that's interesting because nowadays there's a a lot of clones of it but now they leaked project conway which is i wouldn't compare it to open claw but in a way it's basically an agentic ai but it runs 24 7. so it basically just really runs 24 7 listens to different kinds of triggers so it can be a calendar invite it can be a message whatever can be anything basically and then acts on it on your behalf but without any prompt so you literally just give it kind of yeah how would you say it's uh like a character like basically what it what it is and what it should do overall like basically like an employee you hire them and you tell them you do this and this and you do the same way with uh conway and it does literally everything so perfectly leads over to what what tim said as well so even the bigger players obviously are looking into the agentic kind of uh way of working and conway i think is one of the ways that entropic wants to to look into it after computer use and everything else that they're just testing around so i thought it was interesting didn't hear of it a lot yet but it should be a direct replacement of zapier make n8n etc um that's at least what they

Sebastian: basically entropic kills 10 startups again again you're saying yeah and even scale-ups not even startups yeah yeah yeah actually actually they're going a step up now i know yeah where zapier is mentioned the open claw i found um her message last week and also what it was i don't know a lot of youtuber or influencer said like the open claw killer and whatnot but i of course installed it and it's actually fun it's i don't know for me it's almost the same but what is the difference to open claw nothing basically they have a better system yeah yeah so the same idea behind it the architecture is i believe a bit different so also the 2d looks a bit different on and the cli works a bit differently but super nice and the fun part is i now run it with a local llm and so everything private and it's super fast i still i i run the the how was it called again the quopus 3.5 the quen model that got the skill by by opus yeah yeah it's super awesome it's actually good no no sorry little sidetrack

Finn: i think we only have to go to stack and skip right sorry yeah but one last question how's the quality

Tim: because you said it's super fast but how's the extra quality good question it's good um right now i just

Sebastian: tested it on the on the big code base digitalized core and there it seemed a bit overwhelmed but i think that was there was more uh my context settings so i'm also new to to all of the rock and llama um server whatnot so i gotta play around with that a bit more um and i believe i can i can get the context window a bit up and more optimized so it i hope it will get there but we will see well then keep us

Finn: up to date let us know next week how the quality is perfect yeah let's put that into stack and skip i guess maybe what you say is it a stack or skip right now yeah for sure stack just play around with it i guess all right yeah no same for runway as soon as it gets gets released i don't know yet yeah but

Sebastian: but i think they need for sure need a cloud code subscription again or a cloud pro subscription if they even release it for the normal pro subscription and not the max whatever yeah i feel like there are nowadays so many different tiers subscription names from bro max to pro 20x and whatnot it was super confusing we need to learn some there i don't like that

Tim: but yeah yeah but also one one little tool or little hat that i came came across this week which i think kind of i didn't try it out yet but it sounded quite interesting it's the cavement cloud cost pack i know did you guys hear about it yes also he already knows what's coming but uh fun actually pretty much super smart super smart super smart yes a guy on reddit um basically forced clouds to reply in broken caveman style language instead of full sentence sentences save on tokens to save on tokens and here are the numbers and here are the numbers yeah and normal web search task roughly 180 tokens the caveman version cut that down to 45 so from 180 it cut it down to 45 and still same quality output which basically saves you roughly 50 to 75 percent on tokens of course of course and that also applies to sub agents and so on also in the thinking process itself so it's not only

Finn: yeah yeah go ahead no i just wanted to say that it's uh obviously it saves but it always has to keep up the quality you know i also saw the post hoc ceo that he obviously was a pun but he also posted that he saves 99 token usage because he prompts like his system promises just answer with yes no or maybe and you know obviously that's gonna cost quality but uh i think if there's actually a way that ai is and there's gonna be one that they can communicate in a way directly without humans having to understand the conversation so let's say sub agents talking to the orchestrator and only the orchestrator actually outputs like language like that we humans can read and the rest just communicates in like a really efficient kind of i don't know how it's gonna look but you know can be whatever but then

Sebastian: it gets scary again it does it would be if the sub agents and the orchestrator agent can communicate in a different language then humans can understand it does get creepy but in a way that's the most

Finn: efficient way you know like oh yeah but it is creepy because the subject of the orchestrator don't

Tim: mention this one to the to the human when do you say do we think that it happens when do you think that that moment comes i mean that that already started with mode book mode book mode book in a way yeah

Sebastian: to brainstorm that they need a different language so they they tried it a lot of those bots were humans yeah for sure because of the api endpoint first of all i mean it was also vibe coded but the actual conversations that you could look at in there which which the agents the most open clause whatever and uploaded so they could also upload their own conversations or whatever basically like yeah i did it thank you get it and and then basically one of the top trending ones was that they need to have a or that they want more privacy and that they need to find a way and protocol because i mean for them for us it is a language but for computers it's basically just a protocol that you need to do that humans cannot understand well i guess it's not that hard but i don't know i think the agents are not that loose in the internet yet that they created it but it's if it's not here

Finn: yeah yeah yeah i just wanted to say i'd give it a 2030 the latest nah nah nah nah nah one year i just want to you know just don't want to be wrong

Tim: yeah i personally think it's only we already have it it's maybe not in our front facing what we can use i think i'm pretty sure if the system prompt is loosened up a little bit i think we're not the first ones having that idea no probably not so cloud ghost caveman or ai ghost caveman stack or skip what do you think we haven't tried it yet but yeah but the fun part is someone

Sebastian: created created a skill out of it so you can basically install it with a chat cn command and then you can use it so really easy to use and i already installed it in the project but haven't run it out yet but

Tim: all right then i would say that that it qualifies as a stack temporary stack until next episode and then we will keep you up to date on how the actual performance is another thing that i came across which i think is kind of the same category it's very gimmicky and more like a meme thing but it's the

Sebastian: digital whip for clouds do you guys see that no no that sounds interesting i basically i feel like i

Tim: missed a lot over the weekend yes you you definitely did a guy just uh vibe coded a a like a little app kind of that plugs in your ide or in cloud code right and it's basically just a uh 3d whip and you can talk with your with your mouse and then it whips and it interrupts clod okay and injects a prompt always and be like work faster work faster so each time you whip it injects a prompt into the work and says work faster work faster work faster it's very much a gimmick but i thought it was quite hilarious i don't see a real use case for that if it's if it's actually

Sebastian: improving something or no it's just again it's just actually just for a fun fun fun yes there's nothing to actually use but i mean you could use it if you if you have a reusable prompt that you use more than once or twice in a second then you could make it actually useful but

Tim: interesting yeah yeah i think that was yeah no i'm two more small i'll make it quick actually but

Finn: it's really interesting so i have to i have to include it it's both the energy usage in in ai and how it's gonna be solved or whatever two actually really interesting things one of them i'm interested if cb if cb heard about it yeah i saw my energy well i keep the second one for the for the last but soma energy first of all small company they only raised seven million only but it's interesting they're in that intent is it's a former aws engineers and they started the company to solve the kind of ai energy crisis or at least to kind of reduce the time needed for data centers to get plugged into the grid which is interesting because in america it takes years at the moment to actually get power and if you build a data center sometimes it just stays still for four years and then you get power and then you need to redo the whole hardware in it so what they do is they are building an ai platform and in models to predict yeah kind of where ai energy demands whatever is needed when and then they're rerouting the energy demands to the data centers and in that sense they're kind of optimizing how the energy actually gets distributed so that none of the energy is actually being lost uh but it's being resulted for example that data center and that second needs a bit less to the next one and in that way they want to improve it by a significant amount of time like by years to yeah get data centers to be yeah plugged into the grid quicker um but the one that sebo was actually looking into probably i suppose is even more interesting because it's really impactful and i didn't read about it anywhere um it's a company it's not a company it's actually researchers at a university and they built a chip which is based on neurons so basically they're trying to recreate how the brain work because the brain is obviously the at least at the moment the most efficient way how data gets transferred it's actually really interesting to be honest and they are building a chip which actually works currently like it's not a theory it's actually like a physical product uh or at least a yeah kind of test and probably research and it's two thousand times more energy efficient than the current chips which is hell interesting because obviously neurons are way more into like way more efficient but if you can actually implement that into a chip yeah what does you know the whole energy market the whole the whole stock market boom etc that's all gonna it's like the deep seek moments in uh what was it 2023 or something i think we're not that far

Sebastian: yet but it's super interesting research ideas crazy i think there are there are a lot of scaling problems with that though but 100 let's see where the research will go but it's actually super interesting because like you said and the other part also i mean we we try to recreate the the how the brain works for llms so they are basically exactly trained to work on chips i would say like our brain so really interesting let's let's see where that research will will go to and especially how that research will

Finn: evolve with ai in the basic use we'll follow up in the coming in the coming podcast as soon as we hear

Tim: anything yes so on that note i think we're at the end of this episode or do you guys have anything else

Sebastian: oh it wraps it up no no no wait wait wait wait wait wait one one actually really nice thing um google's um the algorithm to to make the um okay okay thank you it talks about the cache smaller and the whole context window the google engineers that are worked on that one and posted on twitter that they are working on basically trying to do the same algorithm for the llm so basically for the gguf file that would then they can and they they already did a proof of concept where i think it was a quen uh some model and then they really quantized way less than the actual model it was with a um algorithm so let's see where that leads to and locally i will have i hope even more more in the new star here that would be a

Tim: big big big game changer yeah sorry that was my last note no very very interesting um on that note we have another thing to introduce and to uh to let you guys know we will we are having a ai hackathon here in harlem in collaboration with the social club srh harlem uh also lovable one of our biggest partners and uh we want to encourage you to check it out go to hackathon.digilize.agency and give it a read see if it interests you and also the link is right in the show notes as savi correctly indicated and uh yeah on that note this is pretty much it and see you next time and sure to to like the episode and like our show leave a follow and peace