deko @andrerl

Tokyo based AI & Software Engineer and Portfolio Manager Focused and specialist on Japanese Economy Tokyo, Japan Joined October 2008

Tweets

2K
Followers

346
Following

3K
Likes

5K

deko @andrerl

57 minutes ago

Some people havent noticed yet but reading well @ArtificialAnlys and all benchmarks of all models is becoming extremely important to decide which models you use in production There are cheap chinese models that depending on the task I would recommend way higher than top models from Anthropic and OpenAI that costs 10x more It really depends on the task you have, even before getting to pricing and scale. There is no more one model fits all Thats a skill that will take time to companies to understand the importance un building production systems I know for a fact by working on the industry that current top software engineers that actually make decisions, wouldn't consider all the available models by reading benchmarks when deciding what to use in a production environment Actually senior engineers for companies by my perception have been the worst to respond to AI. They do the decision, but they dont dedicate so much to AI as more younger engineers that are 24/7 on AI and which would more normally have more spread experience in different models. And organizations wont give younger AI frenzy engineers the opportunity they need. Is a renewal time, AI is changing everything, big companies need to understand it needs to renew their staff as well with whoever is coming strongly in contact with AI. Old managers with old software experience cant do that. But they would just pick the general strongest famous model that the company can pay, allocating the resources they have. That right now can give really way more expensive and worse results Navigating benchmarks and picking right model for a production system is a highly required skill, the market will take long to understand that, but it will go this way

0 0 0 12 0

View Details

deko @andrerl

a hour ago

@FirstSquawk what the fuck is this tweet? is it a news? somebody said that? is your opinion? I have no idea what this account is doing anymore

0 0 3 49 0

View Details

deko @andrerl

3 hours ago

@alexandr_wang Doesnt matter, one has a coding/token subscription the other not Make one and I will promptly take it into consideration

0 0 0 431 0

View Details

deko @andrerl

14 hours ago

@canarinho60043 Bro just check how is allocation rules for gpif. They would only change the rules Right now they are forced to allocate 50% in japan and 50% outside. Of each half bonds, half stocks In the past it has been 100% japan afaik

1 0 0 27 0

View Details

deko @andrerl

16 hours ago

@nicdunz Yeah also for me by the charts I see myself using luna and sol, didnt see much point in terra at all

0 0 0 1K 0

View Details

deko @andrerl

19 hours ago

Oh what? Why nobody commenting on this one? Very nice results seems like, waiting for @ArtificialAnlys

KwaiKAT @KwaiAICoder

a day ago

V2.5 delivers strong results across coding and agentic benchmarks: - SWE-Bench Pro: 65.2 - KAT Code Bench: 53.1 - PinchBench: 94.2 - KAT Claw Bench: 85.5 Built for both long-horizon repo engineering and multi-step agentic workflows.

3 2 29 4K 7

0 0 0 70 0

View Details

deko @andrerl

20 hours ago

@jun_song I love minimax models, use it crazily a lot

0 0 3 391 0

View Details

deko @andrerl

20 hours ago

Enquanto tu fala isso uma manda de cavalos tao passando na sua frente pra investir em acoes na bolsa Enquanto tu olha crypto e reza pra subir vc podia analisar algo que realmente faz sentido, entender o que ta rolando com semicondutores, ter investido e feito 1000% teu capital com algumas acoes por ai Acorda

1 0 3 277 0

View Details

deko @andrerl

24 hours ago

No man, just check out how gpif allocation works. They cant allocate more in japan, they have rules that they need to allocate 50% in japan and 50% in foreigner being 25% stocks and 25% bonds for japan and foreign Even if they want to allocate right now everything in japan they cannot

0 0 0 31 1

View Details

deko @andrerl

a day ago

@zephyr_z9 I just used pro on chat for a few stuff and got to say is absolutely great tho

0 0 1 2K 0

View Details

deko @andrerl

2 days ago

😂😂

Nikita Bier @nikitabier

2 days ago

@finkd Hello Mark, would you like me to enable Creator Monetization on your account

1K 638 22K 1.7M 571

0 0 0 101 0

View Details

deko @andrerl

2 days ago

@trq212 Come on, add some selectable resets as well

0 0 0 108 0

View Details

deko @andrerl

2 days ago

@iuditg @ClaudeDevs Thats so bad whe happens, so unconsiderate. @OpenAI doing so much better with selectable resets

0 0 0 289 0

View Details

deko @andrerl

2 days ago

@ClaudeDevs Its like you benefit people with bad managing quota skills I try to manage the quota for the full time and always get f*cked by this

3 0 6 3K 0

View Details

deko @andrerl

2 days ago

@NVIDIAGeForce We had in my family some before but man, I was so excited when my father got a Geforce 3dfx Voodoo. Legendary! Today I run dual 3090 for my desktop for AI, and want to get a DGX Spark next. We have come such a long way 💚 #CollectRTX #RTXPowerPlay #RTXOn

0 0 0 161 0

View Details

deko @andrerl

2 days ago

@finkd Whatever, api not available in japan, let me know when there is something I can actually use. So tired of this

0 0 0 593 0

View Details

deko @andrerl

2 days ago

@AIatMeta @WorldsByMike @whylifeis4 Not available in japan, just say it clear before. Dont make everybody lose their time

1 1 3 320 0

View Details

deko @andrerl

2 days ago

@CaliAmazin @leadlagreport X keeps showing it on my timeline even after I unfollow so I took a few seconds to comment Also is that all you have to say? Attack me and not what I said? Did I said anything wrong? I dont think I did at all

0 0 0 55 0

View Details

deko @andrerl

3 days ago

@TencentHunyuan does the free endpoint uses our data for training or retain our data? @TencentHunyuan I wanted to give it a try but worried about this