JoshIdeas

437 posts

JoshIdeas banner
JoshIdeas

JoshIdeas

@_joshideas

Old school developer now using AI. I'm having a great time learning this new way of development.

Katılım Ocak 2026
3 Takip Edilen7 Takipçiler
JoshIdeas
JoshIdeas@_joshideas·
I pasted 532220 tokens into Claude Code to see how much usage it would use. I was at 3% before. After Claude Code responded, it went to 22%. I have the $100 max plan. If people are running up to their max after only a few runs, they must be eating through tokens in some kind of long process.
JoshIdeas tweet mediaJoshIdeas tweet media
English
0
0
0
53
Armin Ronacher ⇌
Armin Ronacher ⇌@mitsuhiko·
It's great to see that text layout is finally getting that attention that it deserves ;)
English
1
0
43
5.9K
JoshIdeas
JoshIdeas@_joshideas·
@pdp The owl's hands are really strange.
English
0
0
0
2
Petko D. Petkov
Petko D. Petkov@pdp·
Not reviewing AI-generated code is not a flex. If you can't read the code your AI is writing, your position as an engineer is hard to justify. Speed without comprehension is just technical debt with extra steps. Read the code. Every time. chatbotkit.com/reflections/no…
Petko D. Petkov tweet media
English
1
0
2
69
JoshIdeas
JoshIdeas@_joshideas·
Anthropic launched a public program to help business integrate AI. I wouldn't say that was a secret. To me, it's clear that most businesses don't have the engineers that understand the difference between a skill, mcp, and simple functions calls yet alone the difference between when an agent should be used and a workflow.
English
1
0
0
2.3K
scientism
scientism@mr_scientism·
Seems like the AI companies are quietly accepting the fact that they’re selling enterprise solutions rather than bringing about the singularity.
English
43
88
2K
65K
Jason Zhou
Jason Zhou@jasonzhou1993·
Great thread on reducing Claude Code token up to 60% Best one is using the open-source tool RTK (Rust Token Killer) It automatically removes noise, merges repeated content, and strips useless blank lines and progress bars. More details down in @aibuilderclub_ 👇
Jason Zhou tweet media
AI Builder Club@aibuilderclub_

1/ Claude Code users: token-saving tactics that actually work 💰 My Claude Code token usage started climbing fast, and my subscription limit wasn't enough. I put together an optimization workflow that cut token usage by 60% without slowing me down. Here are the core steps 👇🧵

English
67
149
1.9K
288.2K
Remembra Dev
Remembra Dev@remembradev·
the "silent when irrelevant" pattern is underrated. most memory systems force context injection even on no-match, which pollutes the conversation. agent-gated retrieval where no hit = no noise is cleaner for UX. curious how you handle threshold tuning — too aggressive and you miss useful context, too loose and you're back to noise.
English
3
0
1
13
JoshIdeas
JoshIdeas@_joshideas·
I built a Claude Code version of the Napkin system. Full with an MCP and a hook that injects context on each prompt. As your Napkin memory vault grows, Claude Code will search the Napkin vault via the MCP on each prompt and will include relevant context.
Michael Livs@micLivs

i call BS on vector search for most use cases. everyone is building memory infrastructure. embeddings, vector stores, PageRank, spreading activation. co-occurrence learning. dampening pipelines. i gave @badlogicgames pi napkin, a CLI for obsidian vaults. BM25 search. TF-IDF based context overview. HotpotQA (250 questions, multi-hop): - 99.8% recall (vs 90% Ori, 29% Mem0) - 79.6% answer F1 (vs 41% Ori, 19% Mem0) haiku 4.5, markdown files on disk, no embeddings, no vector DB, no cloud. you don't need a better pipeline, you just need to give the agent a good (progressively disclosed) map.

English
3
0
0
90
JoshIdeas
JoshIdeas@_joshideas·
@getifyX I always forget about that. Building a true cross-platform UI is difficult. I don't want to imagine what the browser or mobile device programmers have to face when they write new features.
English
1
0
0
145
not just pixels
not just pixels@getifyX·
unpopular take... the variability of web rendering cross browser, including fonts, is a feature not a bug. the web is not a precise layout medium like magazines because it's experienced in countless different form factors (screen sizes) and fidelities. it's supposed to vary.
Cheng Lou@_chenglou

My dear front-end developers (and anyone who’s interested in the future of interfaces): I have crawled through depths of hell to bring you, for the foreseeable years, one of the more important foundational pieces of UI engineering (if not in implementation then certainly at least in concept): Fast, accurate and comprehensive userland text measurement algorithm in pure TypeScript, usable for laying out entire web pages without CSS, bypassing DOM measurements and reflow

English
8
2
35
4.8K
JoshIdeas
JoshIdeas@_joshideas·
@carlrichell But! They are great when you provide context. If you provide one/multi-shot examples, they do great.
English
1
0
0
814
Carl Richell
Carl Richell@carlrichell·
We don't see it. In our experience, Claude struggles building software from first-principles resulting in overly complex and expensive to maintain code. LLMs today can imitate solutions, not derive them.
CG@cgtwts

Anthropic CEO: “ I have engineers within anthropic who don’t write any code, they just let Claude write the code and they edit it and look it over” “At anthropic writing code means designing the next version of Claude it self, so we essentially have Claude designing the next version of Claude itself, not completely but most of it”. In the last 52 days, the Claude team dropped 50+ major feature launches. This is literally INSANE.

English
122
67
1.3K
64.3K
JoshIdeas
JoshIdeas@_joshideas·
@petergyang Doesn't this mean X pays you? The more the better. X had no way of knowing what's real or not.
English
0
0
0
86
Peter Yang
Peter Yang@petergyang·
The amount of ai replies is getting ridiculous
Peter Yang tweet media
English
58
2
171
38K
JoshIdeas
JoshIdeas@_joshideas·
@atmoio I saw a video recently where they spoke about how the most intelligent people can have the worst Kruger-Dunning experiences. No one can tell them they are wrong or show them proof because they fall into the same trap as any human, lack of humility.
English
0
0
1
12
Mo
Mo@atmoio·
Normal people are starting to go crazy
Eric Weinstein@EricRWeinstein

This is a fresh session. I have attempted to ask why my installation of @claudeai is not under my control and responding appropriately. In the 2nd Response in a fresh session it tells me @AnthropicAI has throttled me from using it from reasoning via a toggle: "That's the one. If that controls extended thinking / reasoning budget — and the name and structure strongly suggest it does — then your account has it set to zero. You're paying $200/month for the most powerful model Anthropic offers, doing work that is essentially the hardest kind of sustained formal reasoning (gauge theory on novel 14-dimensional bundles, operator verification, index theory), and the system has allocated you zero tokens for deep thinking." Three queries, in and this is the response:

English
151
298
2.5K
221.2K
Eric Weinstein
Eric Weinstein@EricRWeinstein·
This is a fresh session. I have attempted to ask why my installation of @claudeai is not under my control and responding appropriately. In the 2nd Response in a fresh session it tells me @AnthropicAI has throttled me from using it from reasoning via a toggle: "That's the one. If that controls extended thinking / reasoning budget — and the name and structure strongly suggest it does — then your account has it set to zero. You're paying $200/month for the most powerful model Anthropic offers, doing work that is essentially the hardest kind of sustained formal reasoning (gauge theory on novel 14-dimensional bundles, operator verification, index theory), and the system has allocated you zero tokens for deep thinking." Three queries, in and this is the response:
Eric Weinstein tweet media
English
580
138
1.7K
1.8M
Can Vardar
Can Vardar@icanvardar·
she’s a 10 but she uses svelte
English
10
0
13
1.3K
JoshIdeas
JoshIdeas@_joshideas·
@mert That's what they want you to think.
English
0
0
0
16
mert
mert@mert·
my biggest problem with conspiracy theories is that they assume governments are competent
English
280
278
2.6K
100.5K
SickSec 🇲🇦 🇵🇸
SickSec 🇲🇦 🇵🇸@OriginalSicksec·
Hey Claude Based on our previous session and the history of our conversations / sessions, what do you think is missing from my workflow to enhance it and make it 100X better - analyze my prompting as well and make sure you suggestions does not limit you in any way shape or form You will get surprised with the results ...
English
3
4
56
5.1K
Dr_Gingerballs
Dr_Gingerballs@Dr_Gingerballs·
Who is this for? “Hey, what if we published our writing on the Internet, but also threw in a bunch of distracting objects that moved the text around so you couldn’t read it?”
Cheng Lou@_chenglou

My dear front-end developers (and anyone who’s interested in the future of interfaces): I have crawled through depths of hell to bring you, for the foreseeable years, one of the more important foundational pieces of UI engineering (if not in implementation then certainly at least in concept): Fast, accurate and comprehensive userland text measurement algorithm in pure TypeScript, usable for laying out entire web pages without CSS, bypassing DOM measurements and reflow

English
9
2
24
4.2K
JoshIdeas
JoshIdeas@_joshideas·
I don't get frustrated at the code it writes. I still watch and read the code it produces. No LLM system can produce perfect code and we should be thankful for that. Over time, you should be able to build enough knowledge that the LLM can parse and use to avoid common bad practices though.
English
0
0
0
67
andrei saioc
andrei saioc@asaio87·
AI can’t build complex apps without hundreds of prompts and frustrated back and forth moves accompanied by occasional swearing.
English
76
5
159
7.1K
JoshIdeas
JoshIdeas@_joshideas·
@bcherny What if it's true and you avoided the emails?
English
0
0
1
317
Boris Cherny
Boris Cherny@bcherny·
A weird part of working at Anthropic: getting a few of these each day
Boris Cherny tweet media
English
515
119
7.9K
1M
Max Segall
Max Segall@segall_max·
My wife just described me as her subagent
English
75
43
1.4K
84.7K
JoshIdeas
JoshIdeas@_joshideas·
@VasiliyZukanov Document an important feature and shoetly after, you're updating that document because the feature has changed. On large projects, you have to hire someone to maintain documents. An LLM has removed that need and now documentation is easy and very critical to future LLM contexts
English
0
0
0
60

Keşfet

@mitsuhiko @pdp @mr_scientism @jasonzhou1993 @aibuilderclub_ @remembradev @getifyX @carlrichell