Hennie

3.3K posts

Hennie banner
Hennie

Hennie

@andyhennie

Designer turned coder 🦄 Prev Head of Design at @mittanbud

Katılım Nisan 2009
1.5K Takip Edilen1.1K Takipçiler
Hennie
Hennie@andyhennie·
@simonw My biggest issue with swift is the lack of e2e testar that does not hijack the computer in «automation mode». Found a way through that?
English
0
0
0
157
Simon Willison
Simon Willison@simonw·
I've been vibe coding SwiftUI menu bar apps for my new Mac, turns out Claude Opus 4.6 and GPT-5.4 are both competent at Swift programming, no need to even open Xcode! simonwillison.net/2026/Mar/27/vi…
English
45
27
497
28.9K
Hennie
Hennie@andyhennie·
@davidgomes We need benchmarks that measure this. I want to keep my LOC down. No benchmark measure LOCs afaik.
English
0
0
0
59
Hennie retweetledi
David Gomes
David Gomes@davidgomes·
I'm unshipping a feature from Cursor, and I can tell that all the SOTA models are really bad at deleting code. They will routinely: - Leave behind `throw Error("Not implemented")` style things - Want to give users notifications "Feature has been deprecated" - Keep tests for features that don't exists, and write stubs for these features in the tests (instead of deleting tests) - Not find a ton of dead code, and not care too much about deleting it - Leave useless comments about deleted code/functionality I think we need to much improve the RL data with lots of feature deletions, because they have been trained to only generate code, not to delete it.
English
45
19
594
39.9K
Hennie retweetledi
Marcin Krzyzanowski
Marcin Krzyzanowski@krzyzanowskim·
I reimplemented "claude" CLI with codex and gpt-5.4-high. It cost $1100 in tokens, and is 73% faster and 80% lower resident memory during sustained interactive use. It is very easy to reverse claude from npm distribution, then reimplement is 1:1. It is indistinguishable from the Anthropic version to the every header and analytics it send back github.com/krzyzanowskim/…
English
97
55
1.2K
169.5K
Martin Jensen
Martin Jensen@Jensen2k·
@andyhennie Ja, allerede gjort det. (Ref bilde). Men så er jo dette en nyansert greie. Vi bytter f.eks ikke ut HubSpot med det første. Dette her er litt ikke-kritisk tooling som er ganske lightweight som vi ikke har fra før: da er det enklere. Så er det en god endringsagent for å vise AI
Martin Jensen tweet media
Norsk
1
0
1
469
Martin Jensen
Martin Jensen@Jensen2k·
Har hatt behov for Calendly/Cal.com – brukte 3(!) timer med Claude Code på å bygge en klone for TRY. Sanntidssynk mot kalender, Teams, SSO – alt inkludert. Sparte over 1 mnok/år🤯 Har vært skeptisk til "SaaS-døden", men.. kanskje? (Claude Code + Remotion har laget videoen)
Norsk
16
3
62
10.3K
@levelsio
@levelsio@levelsio·
Okay let's see who can reply to this
English
2.5K
17
2.2K
1M
gabriel
gabriel@gabriel1·
hello friends
English
575
3
1.2K
96.6K
Hennie retweetledi
pedram.md
pedram.md@pdrmnvd·
men in their 40s used to have cool midlife crisis but now they just have agentic workflows
English
163
678
7.6K
430.1K
Hennie
Hennie@andyhennie·
@levelsio I could not. Norwegian indie founder (1k followers). What can I do to get inside?
English
0
0
0
78
Hennie
Hennie@andyhennie·
a good way to see how intelligent a coding model, is to look at the kind of questions they ask during plan mode. been juggling between claude code and codex today, even using the same prompt in parallel. and codex "gets it" more often. I asks the right kind of questions. so good!
English
0
0
0
50
Hennie retweetledi
rahul
rahul@rahulgs·
seems obvious but: things that are changing rapidly: 1. context windows 2. intelligence / ability to reason within context 3. performance on any given benchmark 4. cost per token things that are not changing much: 1. humans 2. human behavior, preferences, affinities 3. tools, integrations, infrastructure 4. single core cpu performance therefore, ngmi: 1. "i found this method to cut 15% context" 2. "our method improves retrieval performance 10% by using hybrid search" 3. "our finetuned model is cheaper than opus at this benchmark" 4. "our harness does this better because we invented this multi agent system" 5. "we're building a memory system" 6. "context graphs" 7. "we trained an in house specialized rl model to improve task performance in X benchmark at Y% cost reduction" wagmi: 1. product/ui 3. customer acquisition 4. integrations 5. fast linting, ci, skills, feedback for agents 6. background agent infra to parallelize more work 7. speed up your agent verification loops 8. training your users, connecting to their systems and working with their data, meeting them where they are
English
107
222
3.1K
350.3K
nick
nick@tinyblue_dev·
Welp I did it. Wired up MiniMax M2.5 to my 2x Mac Studios (512GB) with @exolabs -> wired into OpenClaw, works equally well to Opus 4.6, and it's free. I'm dumping all my AI Token Subs - SUPER cool day!
nick tweet media
English
28
14
234
37.9K
Hennie
Hennie@andyhennie·
I agree, but two things: 1. Then I assume you’d also be fine with all your Codex/Claude Code conversations being synced to Conductor’s servers? 2. From a security standpoint, that’s a pretty scary attack surface, since you’re allowing Conductor’s servers to execute commands on your machine through Conductor.
English
1
0
0
718
Hennie
Hennie@andyhennie·
TIL: you can add a newline in codex/Claude code with ctrl+j
English
0
0
0
133
Hennie retweetledi
swyx
swyx@swyx·
seriously @walden_yan cooked, this thing legitimately saves my ass 3-8x a day, and yes it sounds weird that devin can catch devin's own mistakes, but this is basically the equivalent of "sleeping on it" and looking at a PR with fresh/more critical eyes. btw you should also see the "smart friend" pattern he piloted in Windsurf, it is going to be the design pattern for smarter subagents (usually subagents are dumber) that others are only just catching on to
swyx tweet media
Gauri Gupta@gauri__gupta

can’t trust merging code without a Devin review anymore. Claude and Codex are great, but they don’t catch bugs the way Devin does. devin code review is OG. @cognition has built something special.

English
31
10
188
25.1K
Hennie retweetledi
Thomas Ricouard
Thomas Ricouard@Dimillian·
The boss's boss has spoken. We're saved. Again, I could not care less about the payout. But X is my main way of reaching a global audience in iOS/macOS and AI communities. So if this is also affecting reach and siloing communities by geography, this would be a disaster.
Elon Musk@elonmusk

@dvorahfr We will pause moving forward with this until further consideration

English
12
4
99
13K
Hennie retweetledi
@levelsio
@levelsio@levelsio·
Terrible for me because my audience is mostly US But I am always outside US, like now Brazil, and other times Portugal or Thailand This isn't great for digital nomads
Nikita Bier@nikitabier

Starting Thursday, we'll be updating our revenue sharing incentives to better reward the content we want on X: We will be giving more weight to impressions from your home region—to encourage content that resonates with people in your country, in neighboring countries and people who speak your language. While we appreciate everyone's opinion on American politics, we hope this will disincentivize gaming the attention of US or Japanese accounts and instead, drive diverse conversations on the platform. We invite creators to start building an audience locally. X will be a much richer community when there's relevant posts for people in all parts of the world.

English
415
49
2.6K
360.9K

Keşfet

@simonw @davidgomes @Jensen2k @levelsio @gabriel1 @tinyblue_dev @exolabs @elonmusk