Embed Learner

Mechanical engineer turned embedded learner!

🇮🇳/acc

[Embed Learner @EmbedLearner]

Space bunny on @opencode has become extremely slow. My context is of about 470k tokens though, can't help but execute all changes in a single session.

[Embed Learner @EmbedLearner]

@ivishaltejwani Unfortunately Jugaad is taken as something to be proud of in India, but it shouldn't be!

[Embed Learner @EmbedLearner]

@SandeepMall Memes have started means now is the time to invest. When it isn't cool, it is time to go in!

[Embed Learner @EmbedLearner]

@mehulmpt At present the subscriptions are extremely subsidised, once you take away that, local AI would start making sense.

[Embed Learner @EmbedLearner]

@MiaAI_lab Spelace bunny felt like a decent model to be honest on opencode!

[Embed Learner @EmbedLearner]

@opencode WTF you only have 3 people in the team!

[Embed Learner @EmbedLearner]

@i2cjak In 2 months you get opus 5.5 level open model!

[Embed Learner @EmbedLearner]

@ivishaltejwani But why?

[Embed Learner @EmbedLearner]

@opencode So you are also hosting it in the US?

[Embed Learner @EmbedLearner]

@NFT_Chen I am generally more excited about flash models, eagerly waiting for GLM 5.5 flash, but hey what happened to 5.4 flash? Skipped?

[Embed Learner @EmbedLearner]

@x_rahulraj LFP or NMC?

[Embed Learner @EmbedLearner]

It's time to play with "space bunny"!

[Embed Learner @EmbedLearner]

@Siddhant__rao Why not switch to linux + KiCAD?

[Embed Learner @EmbedLearner]

@WescheNex1q Mimo looks cool too!

[Embed Learner @EmbedLearner]

Yesterday while playing with GLM 5.3 flash, got some good results. Model is laconic but does it's job fine. Tried a kind of similar task on Deepseek V4.1 Flash today, it went on long thinking rant and in the end couldn't solve the issue. Kind of disappointing to be honest!

[Embed Learner @EmbedLearner]

Today I exclusively used GLM 5.3 flash instead of Deepseek V4.1 flash on my @opencode go plan. I missed the extreme verbosity of DSV4.1F.

[Embed Learner @EmbedLearner]

@thdxr So all open weight models would be hosted by you in the US?

[Embed Learner @EmbedLearner]

@SiddSpeed @sankitdev Nahh, it's benchmaxxed! But hope one day meta makes it open weight

[Embed Learner @EmbedLearner]

@Siddhant__rao Isn't KiCAD coupled with decent model such as Deepseek V4.1 Flash not enough for replacing Altium?

[Embed Learner @EmbedLearner]

@opencode WTF it's taken more than 10mins for a shitty coding task, whats going on?

[Embed Learner @EmbedLearner]

Trying Union Alpha on @opencode go. Weird model, does not show thinking traces, works in the background a lot. Takes it own sweet time. Latest favourite is Deepseek V4.1 Flash which is blazing fast and works great for my usecases.