Zengineering

My curated list of good follows for GenAI / ML twitter.com/i/lists/165767…

@VicVijayakumar because we don’t empower people to be able to ship end to end. because we lack a strategy and our backlogs are limited.

gonna start seeing this more and more: eng leadership is going to ask why aren’t you shipping more features, we gave you unlimited tokens.

@GergelyOrosz Should we count frontier labs in this list - why or why not

I've yet to see, or hear of a company that is winning against its competition, because said company is spending more on AI tools, or using it better than the competition. Ways I see companies win: - better product - better marketing - cheaper prices - better unit economic - more funds raised etc

@sidin They became agent skill guys

Where are all the second brain guys?

@fredsoda definitely expensive.

big issue with AI is that it is an expensive technology to experiment with at scale within enterprises companies probably have one to two quarters to show material results or i see enterprises cutting back on token spend (one reason why OpenAI/Anthropic are going public soon)

Engineers fluent across all three eras are rare: Software Engineering 1.0 Software Engineering 2.0 Software Engineering 3.0 1.0: building infra 2.0: model training and design 3.0: agent harnesses

@HanchungLee iRev go brrrr

ngmi if complaining about looking at data. welcome to ai/ml.

@puram_politics 🤔

@Samhanknr no. It'd make people live in houses they can afford.

There are bank/finance pages suggesting you should buy a car that's 50% of your annual income. And claim that's conservative. That's BS. You should not buy a car that's any more than 25% of your annual income.

@puram_politics That would put housing out of reach for most in urban areas

@Samhanknr 2x.

@echantech1 Can you explain why your cache hit rate is high ?

Argue with it for about 20-40 minutes about its proposed design, research, tradeoffs, proving spike-and-validate with repro/bash scripts to ensure it’s not hallucinating, how it didn’t catch this issue before, etc. Then I make it generate the plan through Cursor + Codex. Then using my skill to split the plan into parallelizable units of work across multiple SSH machines. The result is I have like a 95% cache hit rate on 11bn tokens a month. My estimate is that this costs $3-5k of compute but I’ve NEVER hit my weekly limits ever with my 20x Codex.

lots of people saying software architecture matters and models need to get better at it but interviews barely get past toy system design. real architecture is hard to make legible and the more legible the test, the easier to hill-climb with RL probably the only real signal is long-running work in public

@puram_politics what about houses

This is for people in their 20s. If you are older than 35, no more than 15%.

@AKHomeproject @Nico43688773 @fredsoda surely we can find some statistics that show this ?

@Nico43688773 @Samhanknr @fredsoda unemployment numbers are insanely cooked, underemployment very very common. Idk how you would even measure the quality of work. Depending on age you see a very different economy, mostly based on when housing was bought (or not)

one thing to note here is that everyone always talks about low unemployment but no one ever talks about the quality of employment most new jobs are either gig economy or healthcare, neither of which pay well when it comes to a high inflation environment

@echantech1 What’s your workflow ?

@Samhanknr AI does a pretty decent job actually with architecture. You might have to argue with it back and forth until you get exactly what you want but if you can get it to justify its answer and explain readers it’s pretty effective

if we see this, we know

@Polit_eurOpines @Nico43688773 @fredsoda okay so does that show up in any statistic ?

@Samhanknr @Nico43688773 @fredsoda the vibecession is because the poorest third of Americans were doing much better while the richest third, who control news orgs, disproportionately donate to political organizations, and manipulate social media, were doing worse

@Nico43688773 @fredsoda my pet theory is that there’s an increased perception of precarity combined with increased competition and comparison. the median incomes maybe going up but also the volatility is high maybe

@Samhanknr @fredsoda i agree that its a vibecession, @fredsoda seems to imply that unemployment is low because of poor quality jobs/gig work. but its not true. my theory is people are exposed to so many more scams now, (crypto, gambling, MLMs, etc) that people feel more drained now than before

@buildwithparas yeah but you need to be able to turn it into a reward function and that’s highly non trivial

@Samhanknr why they picked it usually sits in old prs and chat, the repo just shows what landed

@Nico43688773 @fredsoda that’s why it’s called a vibe cession. you can also check real disposable income. they are also still going up iirc

@fredsoda @Samhanknr median real wages have been going up though. it was moderately stagnant frm 2000-2014 but after 2014 its been shooting up, right when gig work became mainstream i dont think gigwork pushed up wages, but somethiing is pushing up wages more than gigwork is depressing them.

@fredsoda glad we touched base

i work in a corporate environment lmfao, what do you expect

@fredsoda double clicking - you’re picking up corporate jargon

yeah I’ve seen this and double clicking on the internship part it’s critical if you’re in college to start internships early i feel like this wasn’t a thing for tech but now it’s mandatory if you want have a chance to secure post-grad employment comfortably