I recall when I first saw Ubers’s H3[1], it really reminded me of healpix (from grad school). I know the algorithm is different but similar problem space: “given a point on a sphere, assign it to a stable cell ID”
Anyone get a feeling that we are “inventing” a bunch of things that will be subsumed by the model in near(ish) future? Like the whole context refresh thing? Codex is already better (ime) than CC.
And then all these planning steps … if CC/codex interviews a ~million senior devs, next iteration of models will perhaps know how to plan way better than today?
I had a miserable experience with it. It could not do as simple task as discovering my python interpreter! IDEs are a dime a dozen these days — most are fast enough. But common workflows need to be air tight.
This was on nytimes just a few days ago. Quoting the bits that might be relevant[1]. A lot more in tfa
> Those efforts have been complicated by the fact that major hedge funds, including those led by friends of Mr. Bessent, stand to benefit financially from an Argentina economic lifeline. Funds at investment firms including BlackRock, Fidelity and Pimco are heavily invested in Argentina, as are investors such as Stanley Druckenmiller and Robert Citrone, both of whom worked with Mr. Bessent when he was an investor for George Soros.
My beef with TDD is most every resource merely parrots the steps (red,green,..). No one teaches it well from what I have found. Nor am I convinced it’s easy to teach. I have picked up what I can by watching (what I believe) good TDD practitioners.
I have a feeling this is where TDD loses out the most
biggest hope for the survival of human race in the face of an alien onslaught is that we taste like yummy chicken. this way, they will keep us around for a long time
https://x.com/tobi/status/2053121182044451016