Sabitlenmiş Tweet
Maibook v0.1.5 is now out of free preview - Mac and Windows, 16GB RAM. Free to try for 7 days. Get it at maibook.app
English
rcanand
2.6K posts
rcanand
@rcanand
AI Engineer | into local private AI | product/engg/research | #AI #LLMs #Ollama #MLX Built https://t.co/nwAmYD0QJl - local, private, personalized AI agent network.
Bellevue, WA Katılım Mayıs 2008
140 Takip Edilen360 Takipçiler
I am presenting at AI tinkerers Seattle later today.
Building agentic workflows is hard, but the secret weapon is building custom web-based dev tools to visualize and debug what your AI actually sees.
seattle.aitinkerers.org/p/ai-tinkerers…
English
coding is a means to an end, invented because it was a tool
language is fundamentally human
communicating with one another is what makes us different from other animals or machines
one is about efficiency, the other is attempting to imitate millions of years of evolution
rahul@0interestrates
why do people (including me) have an aversion to AI writing but not as much to AI code? if a piece of text smells AI i stop reading it but i use things coded entirely with AI every day
English
yeah, we were sitting at a coffee shop in vienna, and i said: let's write a shitty python web app framework. armin was like "that's a brilliant idea, but i don't know how to program". so i went ahead and taught armin, which explains the python. the rest is history.
Kush@kushbhuwalka
TIL the guys that created pi (which openclaw runs on) created python flask 16 years ago. (@badlogicgames, @mitsuhiko)
English
Been trying to do that - work with coding agents (mostly Claude code) as opposed to having them do the work. One big challenge is that they do many things to take agency away from you. Increasingly, over time. Many examples - no more thinking blocks(most valuable debugging and early steering tool - gone), that stupid hook question at the end (and the choices to pick from with AskUserQuestion which always misses what I wd hv picked) that pushes you where it wants you to go - quietly taking the steering away from you, glossing over the grayest parts when it summarizes what it did (single biggest tell of where it fucked up, every time), writes everything away from your sight, docs initially, tests are python -c transient, never saved, and even worktrees are in .claude folder by default, marking tasks done before my verification, not asking and hallucinating assumptions, often wrong, and many more. Getting harder and harder to do anything beyond just asking it for syntax or to code specific functions or algos it can whip up from memory that wd take me longer. Most of my engineering is fighting its principal level arrogance with engineering skills worse than an entry level dev. The only reason to put up with it is its breadth of factual knowledge, everything else is tax we pay for that. Any tips to get around these?
English
I'm usually not one to write thought pieces without much technical depth. But here we go.
Slow the fuck down.
mariozechner.at/posts/2026-03-…
English
rcanand retweetledi
Disappointing pricing trend with Gemini 3.5 Flash.
22.5x pricier than 2.0 Flash which came out 15 months ago ($9.00 vs $0.40).
Are Flash models supposed to get this much more expensive, or is Pro just being renamed to Flash?
Logan Kilpatrick@OfficialLoganK
Welcome to Gemini 3.5 Flash, our most powerful model to date. It pushes the frontier of intelligence, speed, and cost putting 3.5 Flash in a class of its own. We spent the last 6 months making sure Flash is great for real world use cases. It's available everywhere now!
English
rcanand retweetledi
Gemini Omni doesn't just build scenes that look real, it reasons about what should happen next. It combines an intuitive understanding of physics with Gemini's knowledge of history, science, and cultural context.
Rolling out today starting with video outputs to Google AI Plus, Pro and Ultra subscribers globally through the @Geminiapp + Google Flow, and @YouTube Shorts this week.
English
rcanand retweetledi
Excited to welcome Andrej to the Pretraining team! He'll be building a team focused on using Claude to accelerate pretraining research itself. I can’t think of anyone better suited to do it — looking forward to what we build together!
Andrej Karpathy@karpathy
Personal update: I've joined Anthropic. I think the next few years at the frontier of LLMs will be especially formative. I am very excited to join the team here and get back to R&D. I remain deeply passionate about education and plan to resume my work on it in time.
English
rcanand retweetledi
// Code as Agent Harness //
100+ page report on all things related to agent harnesses.
(bookmark it)
In particular, the survey summarizes methods and applications of code as agent harness.
This paper makes a strong case that code-as-harness might be the key to moving us towards a broader science harness engineering.
Is code all you need? Maybe.
Regardless, the paper argues that future systems must have the following four properties: executable, inspectable, stateful, and governed.
Paper: arxiv.org/abs/2605.18747
Learn to build effective AI agents in our academy: academy.dair.ai
English
rcanand retweetledi
Personal update: I've joined Anthropic. I think the next few years at the frontier of LLMs will be especially formative. I am very excited to join the team here and get back to R&D. I remain deeply passionate about education and plan to resume my work on it in time.
English
rcanand retweetledi
@CodeWithBinary Yeah, I think so too. Many trends in platforms seem to be going in the other direction.
English
Most people are not posting out of any passion, have nothing original to say. For them the whole game is external validation. The rare few who have something original to say are mostly hidden by the algorithm from the rest of the world. Rare people with original thoughts ( like you and a few others ) still reached my feed, so there is still hope :)
English
I think we gotta redefine win here
because even if you hit every prestige metric out of the park and get every external signal of validation, it is such a common trope that this all feels empty and hopeless if you weren't doing it out of your own curiosity and love of the game first and foremost
English
have had a hard time articulating it until now,
but the thing that repels me from some new media accounts..
is that I feel them putting out content from some projection of what they think the masses will like..
but I'm not really interested in that.
I'm interested in people who are putting out content, where they're making the thing to the best of their ability, and they stand by it regardless of the external response, because it is born from their own artistic integrity.
English