0xפנתן‎ (@p_nathan) header
0xפנתן‎ (@p_nathan) profile picture

swe/sre/human I like to program computers and talk to people. Founder of @housecarl_cloud 🧑🏻‍💻/🚲/🏋️‍♂️/🖌️ / ✡️ dms open, 100% of the time.

spokane, wa
Joined January 2010
Photo by 0xפנתן‎ (@p_nathan)Photo by 0xפנתן‎ (@p_nathan)Photo by 0xפנתן‎ (@p_nathan)Photo by 0xפנתן‎ (@p_nathan)Photo by 0xפנתן‎ (@p_nathan)Photo by 0xפנתן‎ (@p_nathan)Photo by 0xפנתן‎ (@p_nathan)Photo by 0xפנתן‎ (@p_nathan)Photo by 0xפנתן‎ (@p_nathan)Photo by 0xפנתן‎ (@p_nathan)
Haven't mentioned Arbiter for a few days. What does Arbiter do? Your bot (Claude, ChatGPT) brings Arbiter portions of documents regarding a question and Arbiter analyzes them. Arbiter is designed to handle contradictions, conflicts, and time variant data.
1
174
The "first generation immune to propaganda" moment
JUST IN: Maryland activist helps block a $5B AI data center planned for the site of an abandoned shopping mall that closed 24 years ago, after gathering over 25,000 signatures.
215
827
55
10,740
289,845
I've got a bright idea
What is an unsexy policy that if passed would immediately help millions of people?
159
940
93
15,869
188,511
AI is made in the image of adam. Perhaps that's all we are, image bearers and endless image makers, replicating the seminal action imprinted into us at creation.
“It just feels like when there's no human in the loop, the agent is incapable of looking at the big picture architecture, and gets so reactionary to every finding - even more so when you have multiple PR bots being triggered in parallel” It’s almost like the “agent” has no agency and is merely an echo of our own human intelligence reflected back at us.
9
'wot is going on here' is a hard problem for an ai to solve when everything is contradictory, vague, misdated, and generally dirty. It's also hard for a human. I built the first version of answering that question with Housecarl Arbiter (see pinned), but its a bigger problem..
1
18
... It's search, but it isn't even that. Dreaming is a hack on the problem, trying to filter the chaos of information through LLMs. A rigorous answer here goes into defeasible logic, subjective logic, para consistent, etc. Worse, it also goes into the hard part of NLP- frame reference, multi document frame reference, etc. Well, I'm working on that whole system for the next version of Arbiter. Arbiter v1 is good. V2 will support a full ingest pipeline to handle these hard topics in a rigorous fashion. I'll be rolling out an beta program here when I have the pipeline working.
8
0xפנתן‎ retweeted
the fact that grass can become a sweater through the intermediary of a sheep ought to astonish us more than it does.
the average person doesn't appreciate wool enough
114
1,699
68
25,520
294,111
this is my expectation of 2028 broadly, and going through 2027 it'll be the standard pattern developing I'm actually pretty bear about frontier labs' future. What exactly should we use opus 8 for when 5.5 does it good enough, and LunaShot 8.675309 matches opus 5.5? (Bedrock, etc standing in for self hosting).
BREAKING: Revolut moved ~2,000 engineers onto self-hosted GLM for coding and pushed Claude Code and Codex to the exception list. Per David Abitbol, who leads AI ops there: 56 Nvidia B200s, a blended 30 cents per million tokens, roughly a tenth of what they paid Anthropic and OpenAI. Engineering lead time improved 70%. Here's the part that matters: the gains showed up when tokens stopped being rationed. Engineers quit saving AI for the hard tickets and pointed it at everything. Frontier models become the exception, open weights become the default. That's a pricing problem for every lab heading to IPO. h/t @OliverMolander Would you move your team off frontier APIs if the bill dropped 90%?
19
0xפנתן‎ retweeted
So much rhetoric about monorepos being bad for agents primarily from people who have no idea how git works clearly. `git log -- folder_in_mono_repo` wow look at that perfect history for only one app in a monorepo. Anyone who has literally ever tried to ship software quickly across a microservice mesh that inevitably arises from separate repos knows how insane this is to choose. "What if deploying required sequencing eight separate steps and required writing code in 8 different places both forward and backwards compatible?" Please. Just use a monorepo and make a monolith, extracting services only when failure characteristics and other boundaries *truly* require it.
who thinks monorepos are good for agents they absolutely are not, it's a completely wrong pattern for agents, it ruins their ability to navigate history because it removes the project specific container there's zero benefit from monorepos now that we have agents, only downsides monorepos were solving a different problem that's no longer even relevant
34
12
7
336
42,440
0xפנתן‎ retweeted
Calling it now. AI is going to usher in a whole new era of software development. I’m calling it Software as a Service. People will want to use these tools but they’ll realize they don’t want to spend the time to maintain and update them over time. This will create a whole new category of businesses and entrepreneurs that can take on that task for businesses that would rather work on their core products. AI is changing everything. This is a whole new paradigm of thinking about the way we interact with software.
i have no time to maintain all my vibe coded tools we need a business model where someone else maintains the code and i just pay a monthly fee
236
233
85
4,418
266,176
0xפנתן‎ retweeted
i have no time to maintain all my vibe coded tools we need a business model where someone else maintains the code and i just pay a monthly fee
560
184
203
4,447
599,402
I have been programming 30 years. I would /love/ to have the capital to fund building out a programming team of these people. The VC that would fund me building a scalable knowledge transfer that gets us to a day 2 ROI positive team would make bank. Number one I need a scaling product I can hire someone and get them to be roi+. Then number two, build the people scaler program. I don't mean consulting. That's a lousy game. I mean really building the thing. Jimmy and Jane from Wotsit, Iowa, 22yo, BS in some field with clear thinking and a drive to work a fair 40-50 hours. I want to turn Jimmy and Jane into builders, people who are not drags on the company. Marketers. Advertisers. Programmers. Sales. That's the Thing. That's the crux of a major problem. We don't have a ramp as far as I can see. I didn't have that problem. I was building stuff (badly) when I was 18 and all I needed was to be aimed. I'm crap at /selling/ - let me be clear. That's my professional challenge for the next decade. But I really thirst to build the scalable "junior positive ROI day 2" pipeline. Because if I can do that and turn the crank? I'll /obliterate/ the competition stuck in their "must hire ultra senior" ways.
This is such a bleak story about what it's like to try to land a job right now. nytimes.com/interactive/2026…
2
61
I believe a major contributing problem here is process-itus and norms that are basically locked into a specific kind of work and flavor of age. Everyone knows you need approvals and process and single sign on and - so forth and so on. That's how you run a company like a proper person. You have crappy enterprise software and stupid levels of cross validation and blah, blah, blah. Really? This is part of why startups win. You can build like 25 year olds with responsibility. Most of the original NASA crew were under 35. You need to have this set of equations: Profit > 0 FCF > 0 (technically you need enough float for the month) Revenue > expenses And you need these people on overhead: - Competent lawyer on retainer to make sure you don't do The Really Dumb (only hire for certain sizes/biz) - Competent HR on retainer to make sure you don't do the Ordinary Dumb (only hire him after a certain size) - Competent accountant to make sure the taxman is paid his due (or for smaller firms: a dedication to reading tax docs). Aight. Now you need to Build And Sell Your Stuff. That's the thing and the whole of the thing. Keep it very very lean. Leaner than you want. For software. Good ideas that aren't used enough: Equity for everyone. They should get upside. And have a clear exit path. Contractors aren't incentive aligned. Avoid for anything but targeted and very finite contracts.
1
10
This is going to be the dominant playbook for a while.
46
AI is a lever for effective delivery by effective people. Why doesn't Microsoft see opportunities to go wild ape vs competitors and system effectiveness with these people?
BREAKING: Microsoft, $MSFT, is offering buyouts to 60,000 workers as part of the company’s efforts to cut costs amid AI adoption, per Bloomberg. Personnel on select teams in the US, Canada, India, and the UK at a vice president level and below will be eligible. This reflects ~27% of Microsoft's total workforce.
44
0xפנתן‎ retweeted
"the perfect sandbox doesn't exi-"
34
199
26
4,205
136,925
0xפנתן‎ retweeted
I swear to god if I hear another person on this app talk about the mainstream media while mainlining the disinfo slop on here all day, I’m going to lose my mind
38
423
20
2,668
66,414
0xפנתן‎ retweeted
Two new Impls. common lisp dot net (clnet) is now hosting container images for two new impls: dotcl and evergreen. Link in comments.
2
4
12
269
Thinking about "memory" in the context of, you know it, AI. Memory is not just curated files. Memory is identifying pertinent facts. Memory is judgement of conflict ( p ^ ~p) Memory is temporal (how does this fact fit into time) Memory is judgement (who should you believe...how much???) But Memory is entity reference and frame coref (who to believe) Memory is analysis (who is colluding) Memory is understanding provenance (who said this and who repeated it). Memory is _online learning _. Ultimately, I think, memory is foundational to having a computer that has a particle of wisdom. I wonder if I could build an AI system that wasn't about knowledge, but wisdom.
1
16
postgres needs an easy solution here for sure.
I've been cooking on this post for a bit, but with the cyber security events of this week I had to ship it: Long-live static credentials must die. We never want to rotate API keys again. Turns out only 8% of cloud and SaaS vendors support secure alternative. This is our call to action to change this AND our commitment to do our part: - We are requiring vendors on our marketplace to support OIDC - We shipped @shadcn components to make it easier to ship OIDC configuration - We are expanding OIDC support on our own platform The last generation got the web from HTTP to HTTPS. It is on us to move past API keys. vercel.com/blog/death-to-lon…
24