adi
@adidoit
Followers
492
Following
12K
Media
312
Statuses
3K
Building @socratifyai - a "personalized Duolingo" for your career goals. Get that job. Get promoted. Get ahead using an AI coach that sharpens how you think.
Toronto, Ontario
Joined December 2008
"Reality has a surprising amount of detail" Evals will get better, real-world behavior will get better, but there's always going to be a gap.
I think the disconnect between the believers and the (reasonable) skeptics is that AI is bulldozing every eval that makes sense but not everyone gets that ease of measurability, demonstrability & learnability are all highly correlated.
0
0
0
RT @METR_Evals: We tested how autonomous AI agents perform on real software tasks from our recent developer productivity RCT. We found a g….
0
43
0
+1 on "You can't properly understand AI progress without using the models for real work every day.". AI is a consumer surplus, bottom-up technology platform. Unfortunately, the late majority still seems to be approaching it with skepticism and then trying maybe one or two.
The vast gulf between all of our actual daily experience with LLM improvements plateauing, vs what some researchers claim is going on, is pretty interesting. You can't properly understand AI progress without using the models for real work every day.
0
0
0
made the same mistake by backing eth over btc in 2017. In retrospect, I was looking for a rational narrative to latch on to with crypto exposure. Luckily, corrected that in the COVID era with more BTC exposure because I started understanding a bit more about memetics.
years ago, i was convinced that eth would flip btc. it made perfect sense: ethereum was fast, scalable, and actually useful. bitcoin was none of these things. i learned a hard lesson that cycle. protocol tokens aren't valued by revenue, P/E ratios, or utility—for a simple.
0
0
1
I keep telling people who zone out about AI that they need to care because the AI narrative is the economy right now. More than consumer, economic sentiment depends on the the belief that GPUs lead to building a lower case god. For better or worse.
Nvidia $NVDA recently reached a nearly 8% weighting in the S&P 500, the highest single-stock weighting since data began in 1981, and surpassing Apple’s $AAPL 2023 peak at 7.24%.
0
0
0
stock vs. flow and all those caveats apply (this figure is a flow not a stock) . But it is a proxy for how GPUs compete with white collar labor at the margin
0
0
0
Very true. At McKinsey if you took the pen (typically the manager level) you had so much leverage on the situation. It's extremely underrated by those who haven't been in strategy formulation, how important certain word choices and phrases become. Ultimately, because you're.
.@pmarca: "The person who writes down the thing has tremendous power.". In most companies, almost no one does it. If you can turn chaos into a coherent plan on paper, people will follow your lead, whether you have the title or not.
0
0
0
Pretty confident all social sites will see higher engagement as agentic coding diffuses. The 60 seconds when your agent is gorging tokens is ripe for a dopamine hit from your favorite brainrot merchant.
0
0
1
via @moretothat. Siddhartha by Herman Hesse has the same point - "Knowledge can be conveyed, but not wisdom. It can be found, it can be lived" . Recommended book. As good as the various essays over the years.
0
0
0
to add to Martin's excellent point - the model makers have (i) access to usage trends (ii) economic incentive to find high margin high volume use cases for tokens. The big use cases are competing with model makers and this will be more acute as models convert . that's why i want.
The most interesting tension in the AI industry:. - better lesson pushes models towards end to end solutions with single large models.- economic viability pushes industry towards many composable models with broad horizontal applicability.
0
0
0
If spec-driven development is the future of software dev then microsoft has to be a big winner. So far it hasn't been the case but Im sure they're eyeing the $200 per month subs and looking for their piece.
arstechnica.com
Microsoft bought GitHub for $7.5 billion in 2018.
0
0
0
interest rate, elite overproduction and AI. makes sense TBH and unclear to what degree each is a driver.
NEW: Why aren't business hiring new grads? Economic consensus has decided it's not AI, but no one has explained what going on. Why young college grads are a risky investment, what Hollywood has to do with it, and how much might be caused by AI. @tryramp data inside.
0
0
2
IMO, if we had abundance but a starry sky I bet we would look to the stars. Unfortunately our abundance is one that closes off the deepest mysteries. Baudrillard was right about simulacra and simulation . .
"It is ridiculous to think that if people worked just 15 or 20 hours a week, they would use their leisure to cut marble or struggle with a musical score. they would brawl, steal, overeat, drink and sleep late."
0
0
0
Still my favorite AI productivity result: . Developers thought they were 20% faster with AI tools, but they were actually 19% slower. Shows why AI isn't a panacea.
We ran a randomized controlled trial to see how much AI coding tools speed up experienced open-source developers. The results surprised us: Developers thought they were 20% faster with AI tools, but they were actually 19% slower when they had access to AI than when they didn't.
0
0
1
RT @SocratifyAI: A core thesis of Socratify is professionals will be good complements to AI. PMs -> learn to code.Devs -> learn business s….
0
1
0
