I do nerdy things with AI, product, culture and code. Building an orchard. I also have rare cars (1:300, 1:750, 1:2282 1:4000) ... that are not worth anything.

Lyons, CO
Joined March 2008
Include

Only show posts containing:

Exclude

Hide posts containing:

Time range
-
Minimum likes
or, or ... the people that kept half assing it and never got the gist and claimed its "death" and uh, yea, are still half assing it. spoilers, most just made it part of how we work ...and moved on, building better and faster
Agile died so fast
20
those Blackstone lab guys must have a sense of humor. I ordered 5 kits ...they sent 10. I'm not even mad.
10
Replying to @mgroves
that comes with a price
23
Replying to @TheJaredPink
maybe their definition of “adult content” is “this will make you feel like an adult?” And jokes on them, I like my torque wrench calibrated wrist. Let the kids figure that out on their own
1
26
I don’t want another do not call list. I want a you’re a scumbag and go to jail list.
85
Replying to @TheHuOfficial
now I'm trying to justify a leather jacket. that thing is fantastic.
35
Wow, what year is it? This still looks like the old YC/Techstars growth-at-all-costs math from the zero-rate years. SCALE SCALE SCALE! At least they have a product that people use, see valuable and pay for. So that’s cool. That $80B net loss …<looks it up> is mostly non-cash accounting on convertibles, so ignore that. Mostly. What’s real is ~$9B of compute spend in against $16.3B revenue, plus hundreds of billions in locked-in take-or-pay cloud and lease commitments over 7–10 years. Those aren’t soft - shortfall payments and non-cancelability on the big Google/Amazon/Microsoft/Broadcom pieces. I’ll keep saying it until people listen - who owns what, where. Spoilers, there it is. Revenue ramp is …wow, which is the only thing that makes the bet …coherent. I’d worry that higher rates just make the fixed-cost overhang worse if growth slows or stagnates. Even worse, becomes unsustainable.
JUST IN: Anthropic's draft IPO filing, per WSJ • Revenue $16.3B in the first half of 2026 • Net loss $80.7B, driven largely by accounting charges • Compute spending nearly $9B in the first half • Cloud commitments about $450B over the next 7 to 10 years • Up to $86B on renting Nvidia chips
32
you don’t need codex for that. Rewind 5, 10, 15 years and more times than I care to admit. “We write unit tests and have over 95% coverage!” That’s great, can I see? Can you run them? <all test pass> Nice …can you change that true to a false <on this critical business function>? <all test pass> “……….um” That should’ve broke something
Codex after finishing: “All 9,741 tests pass.” Great 🥳 btw do any of them test what users actually do?
35
Replying to @HLC_actual
BUUUUHHHH this is fat, BUUHHHH
2
114
Has anyone looked at their paycheck then did the math how much you really take home? I mean after tax, insurance, retirement, and whatever else? Like, what's your REAL take home %?
19
Funny how PRs are becoming the bottleneck. Again. They always have been. They're a smell, not a solve.
8
You're doing it all wrong So many badass, maintainable cars are well under 1000/month.
BREAKING: A record 21.2% of new-car buyers now pay $1,000 or more a month, up from 19.1% a year ago, per CNBC 69% of them took loans of six years or longer to make it work.
53
I see a lot of "WE'RE HIRING" posts and they all feel like engagement farming scams. Has anyone got a single interview from these things or am I calling a spade a spade?
21
have you ran into ...I forget the name of it but its some kind of half-life calculation for decay that you can apply to created (static) vs curated data. Helps a lot in cutting down the embedding noise
19
Replying to @m_zokov
I took years of writing, notes, drawings, transcripts and squashed it down into a rag so other llms can check for "my voice" and to provide points on its style. Works surprisingly well. Same with code examples. try it out - you dont need a huge data set. I think my code one is sub 100k before embedding and my writings/drawings 200mb?
1
97
Replying to @m_zokov
as other have said, its very valueable in domains with good, quality history. Legal, finance, even code bases. When I teach people about it, even with really good prompts/skills/moc/loops/whatever - you’ll always get a strong slant to internet average but putting your data/style/knowledge in front is so much more powerful. My personal favorite as of late - bolting up “tools” (like skills, kinda tiny helper apps) to a RAG with a targeted LLM behind it all.
2
3
5,304
first fire of the season (not 50 today). Also, 70s this weekend.
25
I also fixed a lot under the hood issues (plugs, wires, oil change, gaskets) -- like a rusted cooling system (new radiator, water pump) ...and 7-8 vacuum leaks. Still needs yet another coolant flush, and a full brake bleed... but she runs and drives like a boss.
25