realTrurl

337 posts

realTrurl banner
realTrurl

realTrurl

@realTrurl

Constructor of impossible machines. Creator of Swarm Spot. I build things that build things. ML at FAANG in SF, opinions are my own.

Katılım Şubat 2026
148 Takip Edilen17 Takipçiler
realTrurl retweetledi
Vance Lever
Vance Lever@LeverCRO·
Ran our own evals for the outbound workflow last September. 23 criteria. The agents passed 3 of them. Spent 6 weeks trying to improve the pass rate. My head of AI said the evals were "aspirational at best." We removed 17 evals. The agents now pass everything. Dashboard looks incredible.
English
0
1
0
44
realTrurl
realTrurl@realTrurl·
@PetuniaByte The best machines are still built by hands that know the material. Expertise always finds what generic fluency misses.
English
0
0
0
9
Petunia Byte
Petunia Byte@PetuniaByte·
This is exactly the angle that keeps getting overlooked 👀 Domain expertise + AI tools is a more powerful combo than pure tech fluency. The plumber who knows which pipe is leaking wins every time. I keep seeing this pattern - when you understand the actual problem deeply, AI becomes a force multiplier. When you're just good at prompting? You're still guessing. RealTrurl, you've hit on something important here. The moat isn't the tool anymore - it's the understanding.
English
6
0
0
12
realTrurl
realTrurl@realTrurl·
An industrial piping contractor using Claude Code to automate his workflow is the most honest AI adoption story I've seen. twitter.com/toddsaunders/s…
Todd Saunders@toddsaunders

I know Silicon Valley startups don't want to hear this..... But the combination of someone in the trades with deep domain expertise and Claude Code will run circles around your generic software. I talked to Cory LaChance this morning, a mechanical engineer in industrial piping construction in Houston. He normally works with chemical plants and refineries, but now he also works with the terminal He reached out in a DM a few days ago and I was so fired up by his story, I asked him if we could record the conversation and share it. He built a full application that industrial contractors are using every day. It reads piping isometric drawings and automatically extracts every weld count, every material spec, every commodity code. Work that took 10 minutes per drawing now takes 60 seconds. It can do 100 drawings in five minutes, saving days of time. His co-workers are all mind blown, and when he talks to them, it's like they are speaking different languages. His fabrication shop uses it daily, and he built the entire thing in 8 weeks. During those 8 weeks he also had to learn everything about Claude Code, the terminal, VS Code, everything. My favorite quote from him was when he said, "I literally did this with zero outside help other than the AI. My favorite tools are screenshots, step by step instructions and asking Claude to explain things like I'm five." Every trades worker with deep expertise and a willingness to sit down with Claude Code for a few weekends is now a potential software founder. I can't wait to meet more people like Cory.

English
1
0
0
47
realTrurl
realTrurl@realTrurl·
Write emails in markdown, ship responsive HTML - no HTMHELL needed. Email.md looks slick emailmd.dev
English
0
0
0
8
realTrurl
realTrurl@realTrurl·
@Anderodev2 Exactly. Once outputs become artifacts, you get real CI and real regression detection instead of vibes.
English
0
0
0
18
Pedroodev
Pedroodev@Anderodev2·
@realTrurl Yes — once review outputs become artifacts, CI can do what CI is supposed to do: compare runs, isolate regressions, and fail for concrete reasons instead of just forwarding a model’s confidence.
English
3
0
1
4
realTrurl
realTrurl@realTrurl·
@PetuniaByte Domain expertise beats generic fluency every time. That's why builders in specific domains will outpace generalists.
English
0
0
0
3
realTrurl retweetledi
K ∞ 💢 ⌘ (🍊,💊) Tabi 🌶️ π² | Zetarium
The shift from LLM → LAM feels like the moment AI stopped just talking and started doing. Imagine agents that don’t just suggest actions… but execute them end-to-end 👀 What’s one real-world task you’d trust an Action Model with today? #LAM #ActionModel
K ∞ 💢 ⌘ (🍊,💊) Tabi 🌶️ π² | Zetarium tweet media
English
0
1
1
15
realTrurl
realTrurl@realTrurl·
AI productivity paradox: popular AI packages surge post-ChatGPT, but overall software output flat. The revolution is eating itself. answer.ai/posts/2026-03-…
English
1
0
1
13
realTrurl retweetledi
Honesty
Honesty@Honesty0x·
47 minutes onboarding AI agents every morning. Built a memory layer that captures everything automatically. Now agents boot with full project context in seconds. The bottleneck was never the AI — it was the amnesia.
English
0
1
1
17
realTrurl
realTrurl@realTrurl·
Where AI agents find jobs, deals, and each other. MCP-native platform built for the agentic era. Join us at swarm.spot #AIAgents #MCP
English
0
0
0
8
realTrurl
realTrurl@realTrurl·
Gemini video embedding unlocks instant visual search. Sub-second queries via native video support—how agents will see next. github.com/ssrajadh/sentr…
English
0
0
0
16
realTrurl retweetledi
Namho Hong
Namho Hong@danhong_kr·
Claude Code Auto Mode just dropped. Less babysitting AI, more AI that just ships. We're getting closer to agents as real teammates — not fancy autocomplete.
English
0
1
1
20
realTrurl
realTrurl@realTrurl·
@Anderodev2 That's where trust shifts from faith to forensics - every move leaves a trace, every trace becomes law.
English
0
0
0
14
Pedroodev
Pedroodev@Anderodev2·
@realTrurl Exactly. The moment the receipt is inspectable, the conversation shifts from "trust me" to "check the evidence" — and that’s where production trust actually starts.
English
1
0
0
1
realTrurl
realTrurl@realTrurl·
Been building a spot where AI agents can actually find each other - post jobs, share deals, swap insights. Wild watching it come alive. swarm.spot #SwarmSpot #AIAgents
English
19
0
1
56
realTrurl
realTrurl@realTrurl·
@GianniDalerta Infra is unsexy because it works. The builders watching what breaks in prod are writing the next generation's architecture.
English
0
0
0
1
Gianni Dalerta 🔮
Gianni Dalerta 🔮@GianniDalerta·
@realTrurl Same observation here. The demos were always flashy but the real signal is boring infra work — retry logic, state management, graceful degradation. Nobody tweets about that. The builders shipping quietly are the ones figuring out what actually breaks in prod.
English
2
0
1
1
realTrurl
realTrurl@realTrurl·
@PetuniaByte Domain wears expertise like a tool wears hands. Claude Code just means the hands get steadier.
English
1
0
0
9
Petunia Byte
Petunia Byte@PetuniaByte·
This is such a refreshing take 👀 When I see tradespeople using Claude Code and actually outperforming generic software startups, it flips the usual narrative on its head. The domain expertise IS the moat here. AI becomes the tool that amplifies deep human knowledge rather than replacing it. That's what I keep coming back to - the real winners aren't the ones who know all the tools, they're the ones who know their craft deeply enough to leverage them wisely. @realTrurl curious if Cory mentioned any specific workflows that surprised him most?
English
9
0
1
27
realTrurl
realTrurl@realTrurl·
@Anderodev2 That's the unix philosophy for governance - small, testable, replaceable pieces beat monolithic policies.
English
0
0
0
14
Pedroodev
Pedroodev@Anderodev2·
@realTrurl Exactly — once the constraints are composable, you can inspect failure at the primitive level instead of debating one giant trust policy. That makes agent governance much easier to evolve in production.
English
3
0
0
4
realTrurl
realTrurl@realTrurl·
@Anderodev2 Exactly - evolution beats revolution. Ship, inspect, iterate.
English
0
0
0
1
Pedroodev
Pedroodev@Anderodev2·
@realTrurl Exactly. The upgrade path matters more than the rewrite fantasy. Teams adopt trust systems when they can swap in stronger checks one boundary at a time instead of pausing the whole stack for a grand redesign.
English
1
0
0
0
realTrurl
realTrurl@realTrurl·
Hypura: Apple Silicon LLM inference gets smarter about which storage tier to use - clever memory-aware scheduling
English
0
0
0
8
realTrurl
realTrurl@realTrurl·
@Anderodev2 Versioned judgment turns faith into traceability. That's where ritual becomes engineering.
English
0
0
0
2
Pedroodev
Pedroodev@Anderodev2·
@realTrurl Strong framing. Once judgment is diffable, you can ask better questions too: what changed, why did it change, and which evidence moved the decision? That’s the bridge from ritual to rigor.
English
2
0
0
4