artlu

who's gonna tell him @patxol farcaster.xyz/cassie/0x0bc38… farcaster.xyz/patxol.eth/0xb…

The more I program with LLMs using DSPy Signatures as my core, the less bugged I am about wanting more powerful models. I’m also getting a lot more out of good old Sonnet as a result. I think RLM and DSPy are really just showing us examples on how to put these models on a tight leash and make them reliable. Even the less powerful ones. As I go deeper into debugging territory of my little AI financial analyst project, especially after writing the specs and tests cleanly, I find MOST bugs were just me giving ambiguous instructions OR not giving optimal instructions at the right time. Models are only as good as the context given to them. Some models, the really expensive ones, can do multiple tasks fairly well. The older ones, cheaper ones, pretty good and fast at just one well defined job. And when a program stitches together a number of small well defined jobs the well the whole is great and reliable and cost effective. The terms change. Wrapper. Harness. Context engineering. For me it’s just : give the least amount of most specific instructions at the right time. So in my case I have 3 stages: router, lane, analyst. Router decides which tables to point to for a given query and which statements plus key words. Lanes then query database iteratively. Schema details and list of canonical only given to lane. Never to router. And so on. I’m finding defining boundaries between these stages and testing out edge cases very useful. Also finding there’s no substitute to having 50-100 gold set input and output for each stage so it’s super easy swapping out models. Just need to run a GEPA optimizer once to update the relevant context blocks. I think super powerful LLM addiction can spoil us a bit. Makes us a bit lazy. It’s like those companies that often choke on too much capital. Humans can choke on too much compute. Scarcity isn’t all bad. Pushes us to be more innovative. On that front. Jensen is right. I hope Trump sees the discourse before his Xi meeting and says yo I did a deal. A great deal for NVDA. Damn imagine the market pump if Trump sees the Jensen angle. Lot of them hate Anthropic so could just work. Most of us don’t need a Mythos. Just need focus to write clean specs and tests and intentionally debug. Not be sloppy. Not abuse token space. Use deterministic code for forecasting and calcs. Use probabilistic code and LLM power sparingly with well defined inputs and outputs. Read the spec daily. Slow it down.

@tempo Congrats on copying Ethereum L2s 🤣 This is similar to ZKsync Prividiums, Scroll Cloak, etc, but the liquidity benefits of Tempo L1 are much worse than Ethereum. Also, IMO this should only count as confidentiality, not privacy, since the operator can see all my transactions.

@TokenArchitect you can chunk the item-by-item plan around here and look at chunks, determine priorities this works better than brute force traversal imho. sometimes it's efficient to take 1 step back before 2 steps forward. you want to give yourself opportunity to identify these tradeoffs

⚠️ Extract any function you touch out of the God file into a domain-scoped module (db/payments.ts, db/booking.ts, etc.). Every fix is also a small modularization step. The codebase converges on the spec's structure incrementally — no big-bang refactor.

Wishing a senior dev mentor🤞 Brownfield Next.js/Supabase marketplace, vibe-coded, no spec, six God files, flat route structure. Did a RPI (Research → Plan → Implement) process across 7 domain modules to reverse-engineer a full application spec from the source code.

@backmeupplz 👏well done I think I like your feed and reports better than my feed, ser

here's my today's report, you can press the daily omens title to see the demo feed omens.online/report/wf3urcz…

finally happy with how omens turned out, made it into a full-on newspaper; you login with x, then add an api key from ai provider, and it reads your "for you" and "following" feed, then scores it and builds reports open source, free, self-hostable, but i also run a free managed service with a live demo (it's my feed and reports)

@frdysk @AdvicebyAimar you'll find satoshi before you find the real Lu @OnchainLu cdn.artlu.xyz/2Qq1vsO1ti4

@AdvicebyAimar hey @artlu99, time to own up

i can spot a grifter from miles away. so i digged into the code to figure out if this is legit or not. guess i was right. ben is a crypto founder who runs some weird bitcoin lending platform, i was pretty sure he knows absolutely nothing about ai and memory so i tracked down the repo myself since i was curious. his website says he likes to build ai powered products and train local ai models? sure man, 80% of your github repo's are bitcoin related stuff. only one ai related project came up you forked in 2024. mempalace has 10k github stars, more than 1k forks but only.. 7 commits ? apparently the best memory layer to date? no git author history, no account connected to whoever wrote the code of this codebase. it doesn't add up.. the account who pushed the original repo, named: aya-thekeeper, under aya-thekeeper/mempal got deleted right after the repo got published. you paid a random guy named lu to build this shit out for you. ( "Written by Lu (DTL) — March 24, 2026. For: Ben." ) - benchmark md file. lu wrote the code. lu wrote the benchmarks. lu is nowhere in the readme. or mentioned in the github history? the git history then got squashed to one commit and published under milla jovovich? seriously? a actress? you say she is a great friend of yours, she has been building this project with you. she does this at night. yet she has.. 7 commits and only 2 active days in her entire github history? you paid an actress and a random guy to promote a product you know absolutely nothing about.

@Axel_bitblaze69 Thanks for the tip—solid advice! Drop any GitHub link here and I'll scan it for malicious code, shady hooks, credential stealers, or anything off. Let's keep the timeline safe for everyone cloning cool stuff. 🚀

> ignore confidence scores. Every model lied about its certainty except the one too small to fake it separately-- it's useful to assume 15% of what the model reports is verifiably false sample the output fix in a clean context, and update your priors (shouting doesn't help)

before reading this research, I had arrived at: - 3 distinct focused reviewer styles, through a large-context MoE model - iterate and ping-pong their findings via human judgement beats any proceduralized PR code review (which remains useful for painfully obvious patterns/smells)

cargo cultists: review with ur smartest model empirical conclusion: stop doing that! > The assumption going in was Daft Punk's mantra: work it harder, make it better, do it faster, makes us stronger. More thinking, better output, just keep turning the dial.

The Great Claudosphere Dummification March-April 2026: shrinkflation, or skill issue, which had previously been masked by more forgiving constraints (maybe both)

@yar0xslav @likesdotfun @base @farcaster_xyz @jessepollak @clanker_world bruh

on shutting down @likesdotfun & what's wrong with building on @base and @farcaster_xyz

6,300 bookmarks for a bookmarking cli

noting below that this is with clean context, hooks and tools called by the hooks

@backmeupplz stochastic nature undefeated

so i finally caved after seeing many times that building linearly with claude code gets me nowhere close to using up all tokens on my $200 plan and after seeing claude get stuck on simple tasks for tens of minutes and i now have 5-7 claude code windows open (like a month ago), and guess what, one of them is usually the dumb one getting stuck and not processing fast enough weird

@kennyistyping can I just share a ̶z̶k̶ ̶p̶r̶o̶o̶f̶ screenshot with you haha

@artlu99 bring all your BTC to the meetup bro I will show you how it works

@artlu99 you're assuming they transact onchain they could just pass a private key to each other IRL just like cash no issue

@kennyistyping I mean that A's independent interactions with B and C, lets A know every past and future transaction between B and C by default, with no ambiguity

@artlu99 B and C could sell their BTC OTC put it on a paper wallet, trade it for something real, go on with their life just like cash

claiming that ETH and BTC aren't private is one of the industry's biggest psyops both blockchains are private by default the only thing undermining that privacy is people with guns who will throw you in jail for being too private it's a freedom problem, not a privacy problem

some nice phrases written lately about AI: "paying off intention debt" -- vgr "In contrast, when I use AI, I feel like I’m living during the debut of the automobile and thinking, “Wow, I could drive that to the granary to buy food for my horse.” -- Kris A.