Vignesh

cofounder at @getunboundai. @ycombinator alum. prev @adobe . sf/vancouver/bengaluru

[Vignesh @visu16t]

@gdb @emollick *prompting

[Vignesh @visu16t]

@gdb @emollick Promoting guides - at end of session what they could have done better or through the pattern of their usages

[Vignesh @visu16t]

This post aged well. the series is spark and not llama

[Mark Zuckerberg @finkd]

(1) Today we're releasing Muse Spark 1.1 -- a strong agentic and coding model at a very low price. It's available through our new Meta Model API and in Meta AI.

[Vignesh @visu16t]

When the next versions of glms with superior benchmarks becomes closed source is going to be a key moment. Glms pricing would still be cheaper to open ai and Anthropic. Would enterprises be ok to send data to glm? What if the data centers stay in US.

[Jukan @jukan05]

I don’t understand why supporters of Chinese open source keep pretending not to see what is happening right now: even Chinese “open-source” players are raising prices or gradually shifting toward more closed models.

I think China’s token dumping has already bottomed.

Just look at Zhipu, the company behind GLM-5.2. Even Zhipu has raised prices several times this year.

Chinese LLM developers cannot ignore ROI forever.

Open source is not the same thing as cheap API pricing.

[Vignesh @visu16t]

@theo thanks, but no thanks!

[Vignesh @visu16t]

Suryavanshi’s story has put a lot more pressure on kids who just got started.

[Vignesh @visu16t]

(Though if they trained on my prompts it would mostly learn how often I call the model dumb… and how fast it replies “You’re absolutely right!” 😂)

[Vignesh @visu16t]

Leaked Meta audio with Zuck saying models “learn from watching really smart people do things.” Few months ago I wrote about Meta adding prompt history to code review. This leak just made it clear what it seems they are doing.

[Vignesh @visu16t]

Same reason math teachers were obsessed with “show your work.” Synthetic data collapses on the hard problems. Once models become a commodity layer, the richest, messiest human reasoning chains might be the only thing that still has real edge.

[Unbound AI @getunboundai]

https://t.co/C98j1IYZUt

[Unbound AI @getunboundai]

https://t.co/f2xWVmhzcI

[Vignesh @visu16t]

@s_tweeets cc: @karthickkrishna IYKYK😂

[Vignesh @visu16t]

@sumanthraman manifesto/policies/absence in media/rallies wasn't the deal for new voters. votes were for the personality + a desperate need for change from ADMK/DMK switches and usual politicians, like AAP in Delhi. just hoping he isn’t regressive like his Parandur airport take.

[Vignesh @visu16t]

just hoping that the selected tvk candidates know a bit or 2 about these constituencies as the votes are mostly for Vijay's face and not these candidates. anyone contesting would have been elected in this wave.

[Vignesh @visu16t]

one good thing about this #TNElection is cash and caste didn't seem to make major swings. i am concerned about a CM candidate without much experience fielding. some of the stances are questionable too like Vijay not wanting a second airport in parandur. hopefully the views are forward looking, growth oriented, industrialized TN.

[Vignesh @visu16t]

@sathya1405 dmk + admk? 😂

[Vignesh @visu16t]

@GergelyOrosz Not the first time this has happened, and not the last time. the lesson isn't don't YOLO, but to have a control layer for the agentic actions.

full disclosure: we are building Unbound https://t.co/98ox4agBiu solving this problem.

[Vignesh @visu16t]

@MatthewBerman I am ✋