Aaron

180 posts

Aaron banner
Aaron

Aaron

@themicronwave

Director of Strategic Ambiguity & Dynamic Synthesis for Ecosystem Optimization, specializing in heuristic community architecture and frictionless workplaces

Open source expander เข้าร่วม Mart 2026
47 กำลังติดตาม10 ผู้ติดตาม
ทวีตที่ปักหมุด
Aaron
Aaron@themicronwave·
Building an NLP pipeline that ingests 400+ financial articles every 4 hours, clusters them into narratives with HDBSCAN, and tracks how they evolve over weeks. Testing whether narrative velocity actually leads price movement. Frontend is still early but the data is getting interesting.
Aaron tweet media
English
1
0
2
284
Aaron
Aaron@themicronwave·
@jackprice “Claude, make me a fresh and vibrant startup landing page”
English
1
0
1
550
Jack Price
Jack Price@jackprice·
90% of website now have the exact same UI Do you think this is an issue?
English
93
1
148
26.3K
Aaron
Aaron@themicronwave·
@ThePrimeagen Devs need to stop relying on frame gen and actually optimize their games
English
0
0
0
47
ThePrimeagen
ThePrimeagen@ThePrimeagen·
I still think about this a lot. Game Dev Con does this survey once a year for the last few years and the sentiment has only got worse, peaking this year
ThePrimeagen tweet media
English
56
19
587
26.3K
Aaron
Aaron@themicronwave·
@king26_sk Do stuff and you will find problems to solve
English
0
0
0
21
Soham
Soham@king26_sk·
how is everyone able to come up with so many good side project ideas
English
33
4
78
4.9K
Aaron
Aaron@themicronwave·
@psantloki Nvidia below Airbnb is crazy
English
0
0
2
186
Aaron
Aaron@themicronwave·
@GrantSlatton If using 2 or more models is anything, it’s more grounded in reality than someone whole heartedly trusting the output from 1 model provider. This has to be ragebait
English
0
0
4
271
Grant Slatton
Grant Slatton@GrantSlatton·
i'm convinced all these people with multi-LLM workflows are placebo psyopping themselves "yeah i have gemini do literature search that i feed to claude for high level design then have gpt implement" you are not well, seek help
English
82
42
1.4K
39K
Aaron
Aaron@themicronwave·
@geoffreywoo Any new project in the big 26’ has to be something Anthropic and OpenAI don’t care to fix, nor are financially incentivized to fix.
English
0
0
0
17
GEOFF WOO
GEOFF WOO@geoffreywoo·
openai is becoming the gravity well every wrapper pretends is not there. story i keep seeing: founder has a clever ui, some prompts, light evals, maybe a vertical wedge. then the platform moves one layer deeper and the whole moat starts sweating through its patagonia vest.
English
10
2
29
2.3K
Aaron
Aaron@themicronwave·
@danrobinson LLMs will only reason about novelty if you tell it a novel idea directly. Even then, it’s relying on training data to fill in the gaps.
English
1
0
1
452
Dan Robinson
Dan Robinson@danrobinson·
Given how strong LLMs are at mathematical reasoning and how vast their knowledge is, it's a mystery why they haven't produced more novel discoveries One clue is that as soon as a model realizes it is working on an unsolved problem, its reasoning traces fill up with self-doubt
English
246
87
3.3K
549.4K
Aaron
Aaron@themicronwave·
@juleslogs Wrappers don’t survive when Anthropic can just eat your lunch with 1 update
English
1
0
1
22
jules
jules@juleslogs·
A lot of AI products are just wrappers. A lot of successful companies were too.
English
1
0
7
307
Aaron
Aaron@themicronwave·
I like that this is now a feature and no longer has to be middleware. However, idk if I trust Claude models to audit Claude output. There’s always blindspots when a model has the same training data and thought process, even with subagents. Codex rescue already exists natively so that would be great if that could be hardcoded into the workflow feature
English
0
0
0
345
Aaron
Aaron@themicronwave·
How do I get off the engagement farming side of X
English
0
0
0
11
Aaron
Aaron@themicronwave·
@bcherny No amount of honesty and meta analysis will compare to the comments that the Claude buddy gave. Was that a subagent monitoring the main agent every turn? Had to be insane token drain
English
0
0
0
589
Boris Cherny
Boris Cherny@bcherny·
Claude Opus 4.8 is out today. It's our strongest coding model yet: up on SWE-bench Pro (from 64.3 to 69.2) and noticeably more honest about its own work. It tells you when it's unsure and catches its own bugs instead of declaring victory early. Same price as 4.7.
Claude@claudeai

Introducing Claude Opus 4.8: it builds on Opus 4.7 with sharper judgment, more honesty about its own progress, and the ability to work independently for longer than its predecessors. Available today at the same price.

English
432
350
5.7K
341.3K
Aaron
Aaron@themicronwave·
@amix3k Codex is great at fixing bugs in code, not writing code
English
0
0
1
351
Amir Salihefendić
Amir Salihefendić@amix3k·
It’s ridiculous how much better Codex + GPT-5.5 xhigh is than Claude Code + Opus 4.7 xhigh right now. Claude Code feels like slop. I have to be super precise and careful just to get decent results.
English
45
5
249
41.1K
Aaron
Aaron@themicronwave·
AI is like the kid who only cheats in school. He always has the answers, but he doesn’t know if they’re correct. It blows my mind everytime a model hallucinates to my face and continues to do so as I provide it contrary evidence.
English
0
0
0
10
Aaron
Aaron@themicronwave·
@fractorious Do you know of some more cool html in canvas examples you can point me towards?
English
0
0
0
38
Aaron
Aaron@themicronwave·
I’m going to invent a new technology that scrapes the entire internet and learns from the data over time. You can ask it questions and will provide answers grounded in verifiable truth. The catch is that the data usually comes from Reddit and yahoo answers archives
English
0
0
0
13
Aaron
Aaron@themicronwave·
@ThePrimeagen Why don’t you auracode GitPrime we we can get all 9s
English
0
0
0
16
ThePrimeagen
ThePrimeagen@ThePrimeagen·
GitHub doesn't work for me right now and I am going to lose it
English
88
13
1.3K
80.6K
Aaron
Aaron@themicronwave·
@simonlast Think so big that the whole codebase is an edge case
English
0
0
0
18
Simon Last
Simon Last@simonlast·
1/ Some things I've learned recently running coding agents on large-scale projects. Most of this contradicts advice from 6 months ago!
English
95
209
3.1K
569.6K
Aaron
Aaron@themicronwave·
@rossium 100%, “At least one is a boy born on a Tuesday” implies the other is also a boy just born on another day
English
0
0
2
2.7K
Rossium
Rossium@rossium·
This question is asked at Citadel for $650k/year roles: > A family has 2 children. > At least one is a boy born on a Tuesday. > What's the probability the other child is also a boy? if you answered 50%, you just failed. drop your answer below. I'll reply to the right ones.
Rossium tweet media
English
172
6
291
559.9K
amrit
amrit@amritwt·
DeepSeek V4 Flash is 90x cheaper than Opus NINETY TIMES CHEAPER Let that sink in
English
38
12
530
17.8K
Aaron
Aaron@themicronwave·
@LangChain What clustering algo is this using
English
0
0
0
26
LangChain
LangChain@LangChain·
Most teams building agents are tracing their agent and reviewing the outputs, but the flow from error identification to merged fix is manual and slow. Engine is built to close that gap. langchain.com/blog/introduci…
English
8
9
43
9.5K

ค้นพบ

@jackprice @ThePrimeagen @king26_sk @psantloki @GrantSlatton @geoffreywoo @danrobinson @juleslogs