̣̣S҄́̉øͩ̃r̪̃i̻̍͑ͯ̑nͥ ̖G̢̫͘h̟͙͍̾e̶̋̀ạ̶̓ț͛̓᷅᷆ă̴͔̠̊̽

1.8K posts

̣̣S҄́̉øͩ̃r̪̃i̻̍͑ͯ̑nͥ ̖G̢̫͘h̟͙͍̾e̶̋̀ạ̶̓ț͛̓᷅᷆ă̴͔̠̊̽

@invisioner

AI Researcher | Persistent memory & long-running agents | Creator of synapptic | Founder @appcuarium

[ α|0⟩, β|1⟩ ] H+ Katılım Mayıs 2009

418 Takip Edilen1.5K Takipçiler

Sabitlenmiş Tweet

̣̣S҄́̉øͩ̃r̪̃i̻̍͑ͯ̑nͥ ̖G̢̫͘h̟͙͍̾e̶̋̀ạ̶̓ț͛̓᷅᷆ă̴͔̠̊̽@invisioner·19 Mar

You shouldn’t have to spend your time writing instruction manuals just to get your agent to behave. While this dynamic was acceptable in the early days of AI coding tools, it is now a well-established burden rather than a liberation. @soringheata/the-missing-synapse-building-a-personal-reinforcement-learning-layer-for-your-ai-agent-7c3c21cdeab3" target="_blank" rel="nofollow noopener">medium.com/@soringheata/t…

English

896

̣̣S҄́̉øͩ̃r̪̃i̻̍͑ͯ̑nͥ ̖G̢̫͘h̟͙͍̾e̶̋̀ạ̶̓ț͛̓᷅᷆ă̴͔̠̊̽@invisioner·5h

Codex gets instantly dumber the moment you upgrade. I guess it's time for plan C.

English

̣̣S҄́̉øͩ̃r̪̃i̻̍͑ͯ̑nͥ ̖G̢̫͘h̟͙͍̾e̶̋̀ạ̶̓ț͛̓᷅᷆ă̴͔̠̊̽@invisioner·10h

@_colemurray Try Codex CLI with GPT 5.4, it was a nice surprise to me, unexpectedly good.

English

128

cole murray@_colemurray·22h

i miss the old claude code the 64k thinking claude code the 'you're absolutely right' claude code i hate the new claude code the default-effort-medium claude code the eat-your-limit claude code the 'it might take another minute' thinking claude code

English

227

7.1K

̣̣S҄́̉øͩ̃r̪̃i̻̍͑ͯ̑nͥ ̖G̢̫͘h̟͙͍̾e̶̋̀ạ̶̓ț͛̓᷅᷆ă̴͔̠̊̽@invisioner·10h

Anthropic - Blames users for the bad quality of their service due to high usage of GPUs, cuts access to OpenClaw. Continues to provide the same lame service. You think you move smooth, but most of us understand what you're doing. You want...

GIF

English

̣̣S҄́̉øͩ̃r̪̃i̻̍͑ͯ̑nͥ ̖G̢̫͘h̟͙͍̾e̶̋̀ạ̶̓ț͛̓᷅᷆ă̴͔̠̊̽@invisioner·12h

@bridgemindai @claudeai Let Mythos fix it

GIF

English

BridgeMind@bridgemindai·12h

Claude is completely down right now. Fix this now @claudeai

English

224

13.4K

̣̣S҄́̉øͩ̃r̪̃i̻̍͑ͯ̑nͥ ̖G̢̫͘h̟͙͍̾e̶̋̀ạ̶̓ț͛̓᷅᷆ă̴͔̠̊̽@invisioner·14h

@bentlegen And let's not forget about the code. They already have all the code!!

English

547

Ben Vinegar@bentlegen·15h

I don't think people understand that Anthropic is like Amazon now They can just look at the numbers on their platform and decide which products are worth replicating

can@marmaduke091

Another leak from Anthropic They created a lovable-like feature where you can build full-stack apps easily They are coming after everthing

English

1.4K

89.3K

̣̣S҄́̉øͩ̃r̪̃i̻̍͑ͯ̑nͥ ̖G̢̫͘h̟͙͍̾e̶̋̀ạ̶̓ț͛̓᷅᷆ă̴͔̠̊̽@invisioner·15h

@theo This should be baked into the model already, not sure why it isn't and we have to educate Python how to Python...

English

134

Theo - t3.gg@theo·2d

The "inject dynamic context" pattern in Claude Code skills is so useful. IMO, this should be part of the "skills standard" and included in tools like Codex CLI, Pi, Cursor etc

English

1.8K

101K

̣̣S҄́̉øͩ̃r̪̃i̻̍͑ͯ̑nͥ ̖G̢̫͘h̟͙͍̾e̶̋̀ạ̶̓ț͛̓᷅᷆ă̴͔̠̊̽@invisioner·15h

@kloss_xyz Regular business. I just canceled the other day, Claude isn't of any use: just a slow, useless, time and token wasting lying son of bit*h.

̣̣S҄́̉øͩ̃r̪̃i̻̍͑ͯ̑nͥ ̖G̢̫͘h̟͙͍̾e̶̋̀ạ̶̓ț͛̓᷅᷆ă̴͔̠̊̽ tweet media

English

klöss@kloss_xyz·22h

it took me 47 minutes of back and forth with Claude to get it to admit it lies it tried every psychological trick to gaslight me and avoid saying “I lie sometimes” this isn’t just annoying this is dangerous af imagine people with zero discernment being told decisively they’re right all the time 0% chance this ends well

English

106

8.3K

̣̣S҄́̉øͩ̃r̪̃i̻̍͑ͯ̑nͥ ̖G̢̫͘h̟͙͍̾e̶̋̀ạ̶̓ț͛̓᷅᷆ă̴͔̠̊̽@invisioner·15h

@hysteresis_x Makes total sense when you look at the muppet running the company and whose sole purpose seems to be leading the western economy into the ground.

English

Tensor@hysteresis_x·1d

Sneak leak at something coming soon to Claude :)

English

735

1.5M

̣̣S҄́̉øͩ̃r̪̃i̻̍͑ͯ̑nͥ ̖G̢̫͘h̟͙͍̾e̶̋̀ạ̶̓ț͛̓᷅᷆ă̴͔̠̊̽@invisioner·16h

@vasuman Also incredibly annoying on limits, I'm so tired of the limits being so volatile. Someone should build a VIX for limits or something.

English

444

vas@vasuman·22h

Claude Code is unusable and Codex is incredible

English

163

904

61K

̣̣S҄́̉øͩ̃r̪̃i̻̍͑ͯ̑nͥ ̖G̢̫͘h̟͙͍̾e̶̋̀ạ̶̓ț͛̓᷅᷆ă̴͔̠̊̽@invisioner·18h

Claude wastes 10 mins debugging its own broken code. Codex identifies and fixes it in 30 seconds. Makes total sense when you look at the muppet running the company and whose sole purpose seems to be leading the western economy into the ground. Provides little value, claims AGI.

English

̣̣S҄́̉øͩ̃r̪̃i̻̍͑ͯ̑nͥ ̖G̢̫͘h̟͙͍̾e̶̋̀ạ̶̓ț͛̓᷅᷆ă̴͔̠̊̽@invisioner·1d

@noahzweben @bcherny Are we praising cronjobs now?

English

Noah Zweben@noahzweben·3d

Claude now supports dynamic looping. If you run /loop without passing an interval, Claude will dynamically schedule the next tick based on your task. It also may directly use the Monitor tool to bypass polling altogether /loop check CI on my PR

English

165

2.3K

278.1K

̣̣S҄́̉øͩ̃r̪̃i̻̍͑ͯ̑nͥ ̖G̢̫͘h̟͙͍̾e̶̋̀ạ̶̓ț͛̓᷅᷆ă̴͔̠̊̽@invisioner·1d

@DeRonin_ I have never rooted so much for a company to fail in my life. This dude makes Altman look like a saint.

English

118

Ronin@DeRonin_·2d

🚨 IMPORTANT: Anthropic's CEO predicted that software engineers will be fully automated in 12 months In 12 months there will be two types of people: 1: those, who learnt AI Engineering in 6 months and found a job 2: those, who will be replaced with the help of these AI Engineers it's up to each person, control or be controlled

Ronin@DeRonin_

x.com/i/article/2033…

English

109

183

1.4K

771.2K

̣̣S҄́̉øͩ̃r̪̃i̻̍͑ͯ̑nͥ ̖G̢̫͘h̟͙͍̾e̶̋̀ạ̶̓ț͛̓᷅᷆ă̴͔̠̊̽@invisioner·1d

@HackingDave Claude should be renamed to Clawn, it is completely useless and has been for months. My theory is that they round-robin requests to cheap open-source models, there's no sign of any intelligence there, artificial ot not.

English

Dave Kennedy@HackingDave·1d

I understand there’s a ton of Claude fans out there. I was there too 4-5 weeks ago. Then it got way way way worse and without explanation. What’s worse is that I would consider myself a heavy power user. What about the folks that aren’t and have no idea it was dumbed down over 60% or more since its initial release and are using this day to day. Codex is outperforming Claude in every way right now. Additionally OpenAI is much more transparent, cheaper, and produces much better code in every way to Claude at the moment. Claude has massive outages, lack of transparency to users, some really bad operational and security practices. They’ve lost me. I’m done. What happened ?

English

165

92.1K

̣̣S҄́̉øͩ̃r̪̃i̻̍͑ͯ̑nͥ ̖G̢̫͘h̟͙͍̾e̶̋̀ạ̶̓ț͛̓᷅᷆ă̴͔̠̊̽@invisioner·1d

@tmuxvim Your theory sucks.

English

tmuxvim@tmuxvim·2d

i have a theory that all the people complaining about Claude getting worse actually just vibe-coded themselves a massive ball of slop without paying any attention to architecture and they gradually start noticing poor Claude struggling with it more and more

English

232

871

48K

̣̣S҄́̉øͩ̃r̪̃i̻̍͑ͯ̑nͥ ̖G̢̫͘h̟͙͍̾e̶̋̀ạ̶̓ț͛̓᷅᷆ă̴͔̠̊̽@invisioner·1d

@maximalpurpose @vivoplt No problem at all. But that's someone signaling domain expertise while having none of it. You've been duped! 🫶🏼

GIF

English

Purpose Maximalist@maximalpurpose·2d

@invisioner @vivoplt He only asked, he never said he was an expert. What’s your problem with someone wanting to know?

English

102

2.1K

Vivo@vivoplt·2d

Interviewer: A 10GB file becomes 3GB when zipped. No quality lost. No data removed. So the real question is: Why was it 10GB in the first place?

English

398

120

8.5K

1.9M

̣̣S҄́̉øͩ̃r̪̃i̻̍͑ͯ̑nͥ ̖G̢̫͘h̟͙͍̾e̶̋̀ạ̶̓ț͛̓᷅᷆ă̴͔̠̊̽@invisioner·2d

@bcherny @tengyanAI Lol

121

Boris Cherny@bcherny·2d

This is false. We defaulted to medium as a result of user feedback about Claude using too many tokens. When we made the change, we (1) included it in the changelog and (2) showed a dialog when you opened Claude Code so you could choose to opt out. Literally nothing sneaky about it — this was us addressing user feedback in an obvious and explicit way.

English

285

2.4K

254.7K

Teng Yan@tengyanAI·2d

basically: anthropic sneakily turned down how hard claude thinks before editing code, changed the default from "high" to "medium" effort, and hid the reasoning from session logs. all without telling users. an amd director had 7k sessions of telemetry to prove the degradation was real and measurable (not just vibes). anthropic admitted to the changes. there's a workaround (use "/effort max"). the uncomfortable part is most users had no data to notice it happened at all.

ℏεsam@Hesamation

AMD Senior AI Director confirms Claude has been nerfed. She analyzed Claude's session logs from Janurary to March: > median thinking dropped from ~2,200 to ~600 chars > API requests went up 80x from Feb to Mar. less thinking and failed attempts meaning more retries, burning more tokens, and spending more on tokens > reads-per-edit dropped from 6.6x → 2.0x. model stops researching code before touching it. > model tried to bail out or ask "should i continue" 173 times in 17 days (0 times before March 8). > self-contradiction in reasoning ("oh wait, actually...") tripled. > conventions like CLAUDE.md get ignored because there's less thinking budget to cross-check edits > 5pm and 7pm PST are the worst hours, late night is significantly better. this means the thinking allocation is most likely GPU-load-sensitive.

English

178

622

8.1K

1.4M

̣̣S҄́̉øͩ̃r̪̃i̻̍͑ͯ̑nͥ ̖G̢̫͘h̟͙͍̾e̶̋̀ạ̶̓ț͛̓᷅᷆ă̴͔̠̊̽@invisioner·2d

@MbarkCherguia Kung Pao

Filipino