Rajesh Vasa

Helping organisations leverage AI. Head of Translational Research, Deakin Applied AI Initiative. Robustness, Complexity, Evolution, Knowledge Systems, Advaita.

[Rajesh Vasa @rvasa]

Interesting observation. Given humans are ultimately users of the systems & we still need to do stuff that is interesting, useful, entertaining for humans .. maybe human work moves to discovery of spec. I am with @svembu though on that code gen will eventually be AI & that too, much of the gains will be post moving humans out of the loop as sutton noted. My timelines are a transition over decades (not few years) .. some sectors moving faster than others. Econ & social impact for those of us in tech is baked in and better to prep to avoid fate of Nokia.

[Rajesh Vasa @rvasa]

@MatthewBerman @bot If still open. Me interested too.

[Rajesh Vasa @rvasa]

@rakyll The more a component is reused the more asaptation pressure it faces to change (to a limit). Bugs, new features, external env changes. Ironically, popularity adds more change pressure too. Lets see how it plays out as we accel with coding agents

[Rajesh Vasa @rvasa]

How about creativity as freedom from past (of the agent observing). Ofcourse the observer needs a reference and sufficinet overlap to even understand .. else it can be iust noise or beyond sensing range. Given the spaces these LLMs are operating ... they absolutely can keep making stuff that would appear to be creative to many of us humans.

[Rajesh Vasa @rvasa]

@ihteshamali Thank you for writing such a concise summary. Always good to be reminded of the essence. One of those books that never quite leaves your head fully (some essense always lingers).

[Rajesh Vasa @rvasa]

@leecronin Creativity is freedom from the past in all gradients/abstractions in problem & solution spaces. A perfect intelligence is generating creative outputs to others (once we absorb the new idea/knowledge .. it is now in the past; no longer deemed 2b creative).

[Rajesh Vasa @rvasa]

@Dan_Jeffries1 @Sergey_lll There is suboptimal use of human potential & misallocation; this will be disrupted due to ambient AI in the short-term. Ppl need not panic, we will just be suboptimal at the higher levels of abstraction 😀

[Rajesh Vasa @rvasa]

@sd_marlow @MLStreetTalk @JrKibs Creativity is freedom from past. Eye of beholder & each with their own history. In engineering, physics of domain applies; in art we have an open canvas. Enjoyed perspective.

[Rajesh Vasa @rvasa]

@nbaschez In that case, looking forward to the tool / playing with Tolaria but I keep going back to Z, VSCode.

[Rajesh Vasa @rvasa]

@rohanpaul_ai Memory biases future actions. Also, memory retrival is reflexive in practice (based on what agent sees). Retreival efficacy is linked to goal (and the plan to get to goal is in memory). Any change impacts on all of above. Maybe AGI is not just around the corner :)

[Rajesh Vasa @rvasa]

@anirbanbandyo The tweet seems to have been cut midway & the 12 protocols link did not get posted.

[Rajesh Vasa @rvasa]

@rohanpaul_ai I did a longer essay on this issue a few months ago incl. philosophical grounding with similar cautions. https://t.co/eIiIf6iUpH

[Rajesh Vasa @rvasa]

@bakkermichiel @brianchristian @emollick @AlexTamkin @judyhshen @jankulveit @raymondadouglas @DavidDuvenaud The work was very interesting. I did a longer essay on this from a more philosophical argument in Feb'26; updated it with your work & it helped me deepen my core arguments. Thank you. https://t.co/eIiIf6iUpH

[Rajesh Vasa @rvasa]

"The deepest risk of the current trajectory is not that AI will fail to replicate expert knowledge. It is that AI will succeed just enough to disrupt the conditions under which expert knowledge forms, while failing to provide an adequate substitute." https://t.co/NCVrQE5GD2

[Rajesh Vasa @rvasa]

@martinmrmar Commentary by Swami Sarvapriyananda on Bhagavat Gita (~ helps having the extra elaboration as the context is set better). These videos take a non-dual position (ISKON takes a dualism perspective). He explains the differences too. https://t.co/frBtDdcsDi

[Rajesh Vasa @rvasa]

@tankots @WisprFlow Congrats. Enjoyed the vid for Wispr Flow. Was it made with AI. Also, is there a way for your tool to tell me if the shotcut key I want to set is taken by another tool?

[Rajesh Vasa @rvasa]

@sarthakgh A longer meandering essay about the phenomenon at https://t.co/eIiIf6iUpH

[Rajesh Vasa @rvasa]

@simonw Absolutely 100%. Observed same & got burned on real systems due to this. A question - what tools, workflows and systems will encode in us sufficient theory of the systems we build to overcome cognitive debt? And are OAI and claude eng teams accumulating this high interest debt?

[Rajesh Vasa @rvasa]

@svembu Another one on experts still being needed https://t.co/ufnvvz297F & who will capture the value. also https://t.co/vhE2B0Avm4

[Rajesh Vasa @rvasa]

@svembu @svembu I have been exploring these ideas a lot on my essays; my reflection on agency that AI enables + linking it to Bhagavat Gita at https://t.co/s9FoXjxZhQ