wthw

SWE / python, rust, swift, herdr, LLMs, TUIs, tiling WMs, linux ricer, libertarian, keyboards, pixel-art, low-profile, manga and Bitcoin. mainly RTs and replies

[wthw @0xwthw]

@danc_danc @joshmatz @theo I'd say if codex cli has the same voice mode as in the desktop app, that'd be basically heaven. codex computer-use is already top-notch, if you can remove those annoying MCP warnings from the beginning and put something more compact like opencode does that would be nice too.

[wthw @0xwthw]

@dhh There’s a beautiful projected called spoTUIfy out there and I love it

[wthw @0xwthw]

@Bonkytheidiot @Cr7Godbrand We all do, but it doesn’t mean they’re not like that, it means you’re not their target

[wthw @0xwthw]

@chongdashu I've hooked up local models with hermes via llama.cpp, it's slow still, but worth giving it a try

[wthw @0xwthw]

- pi: best harness according to the benchmarks
- opencode: by far the best UX
- codex: has hard to beat features SOTA computer-use + $imagegen via gpt-image-2

[wthw @0xwthw]

@steipete can't wait for people start building tools to counter this bullshit, even funnier when it's built by open models.

[wthw @0xwthw]

@rikkoo0 @tristansautism @krishdotdev of course we will, and we'll use open models to build tools to counter this bullshit.

[Vansh @vansh22b]

Reasons to cancel your anthropic/claude subscription

1) bad limits
2) bad UI for claude code
3) fear mongering
4) fable limited use in subscription
5) bad harness
6) high downtime
7) not allowed with openclaw
8) not allowed with other harnesses
9) mark on your outputs
10) too expensive
11) hate dario
12) strict filters
13) fable restricted for various use cases
14) useless sonnet and opus models
15) haiku discontinued
16) no support on X
17) no human to interact with unlike tibo from openai
18) recent quality drops
19) strict rate limits
20) poaching top talent
21) hates open source ai
22) anti human
23) believes that they are Saint
24) slow
25) ban accounts anytime
26) heavy guardrails
27) bad cowork
28) expensive
29) tibo is not in anthropic
30) no limits resets
31) complete closed source
32) no grok like humor
33) bad at writing
34) bad at legal
35) bad at cyber security
36) x silence from teams
37) sucks at ui
38) artifacts half baked
39) cult like culture
40) buggy
41) better alternatives exists
42) treats adults like children
43) aggressive content filters
44) sudden model depreciations
45) ignores users
46) centred around enterprises
47) anti open source
48) do not believe us
49) limits don't last enough
50) woke models
51) models are far left
52) do not seek truth
53) do not make models available for masses
54) scraped internet
55) better value from chinese labs
56) better value from openai
57) external agents cut off
58) extra usage pricing insane
59) rug-pull pricing changes
60) source code leaks ignored
61) charged after cancel attempts
62) google play refund issues
63) watermark on output
64) bans and agent blocks
65) no care for sub users
66) focus elsewhere than product
67) endless bug-fix loops
68) constant apologizing instead of fixing
69) opus 5 letdown
70) no acknowledgment of regressions
71) tracking users via outputs
72) no consent for identity linking
73) control sold as safety
74) hypocrisy on AI risk vs profit
75) moral grandstanding while scaling
76) treating users as threats
77) watermarking human-edited text
78) claiming IP over user work
79) training on public data then marking outputs theirs
80) only Anthropic can read the marks
81) dual-layer text + C2PA tracking
82) doom marketing while scaling hard
83) patronizing moral lectures
84) ethics narrative as pure PR
85) fascist-style control mindset
86) guardrails that poison useful output
87) no public detector for their watermarks
88) forcing AI authorship credit
89) enclosing the cultural commons
90) repressive corporate playbook
91) moral borders as soft censorship
92) lecturing instead of clean refusal
93) undermining user ownership
94) global surveillance marks by default
95) compliance theater beyond legal need
96) co-opting user creativity
97) invisible ownership grab on edits
98) data-theft origins of the model
99) AI-welfare talk as cover
100) excessive paternalism toward adults

I think this is enough.

Just cancel your claude subscription guys.

[wthw @0xwthw]

@jordanurbs Really well said.

[Jordan Urbs @jordanurbs]

Just cancelled my Claude account.

It was a fun 2 years watching Anthropic create an excellent product... but at a certain point a man's got to look himself in the mirror and ask why he's paying a company he disagrees with on so many levels.

To replace this gap in my life, I'll be upgrading to the $200 Venice Max plan, to build mainly with Kimi, GLM, and Deepseek. I'll have less tokens to waste but at least my money will go towards a mission I believe in... Vote with your dollar, as they say.

I'll post updates to share how it goes.

[wthw @0xwthw]

@arthurkatcher https://t.co/LZNpyuNMuP

[wthw @0xwthw]

@jun_song Slopthropic

[Arthur Katcher @arthurkatcher]

Anthropic is on a loosing streak.

- Watermarked text & code
- Sonnet 5 is just a bad model
- Fable 5 barely can be distinguished from Opus 5, after 20-30% context load it becomes extremely dumb
- Just bad vibes on X lately.

Should I switch to Codex?

[wthw @0xwthw]

@techdrivenpetru @jun_song so the whole world has to pay for EU bullshit?

[wthw @0xwthw]

@matviy @jun_song https://t.co/LZNpyuNMuP

[wthw @0xwthw]

@jun_song Slopthropic

[wthw @0xwthw]

@jun_song Slopthropic

[wthw @0xwthw]

@jjacky you're in a bubble of the bubble

[wthw @0xwthw]

@jayair that's the new marketing style i guess huh

[wthw @0xwthw]

@pcshipp @i_mika_el the slop is hurting your conversion

[wthw @0xwthw]

@Lifeincarnatex mão de gata mô

[wthw @0xwthw]

@trq212 Another Slopthropic banger.