Brad Campbell
42 posts
Brad Campbell
@uberboringguy
An uber boring guy, but you knew that already. Building fun tools @ https://t.co/Wph5a67Lcq Ham callsign: N4SRF.
Katılım Şubat 2024
37 Takip Edilen2 Takipçiler
@uberboringguy Hi Brad, we’re sorry you’re having trouble accessing your account and understand how frustrating it is to hit a roadblock when seeking support.
Send us a DM with the email associated with your account and company name so a member of our team can reach out!
English
@dialpad I need some help getting into our account. Unfortunately, to even submit a CSR it appears you need to be able to log in to the Dialpad account. Happy to move to DM to resolve.
English
@DashPayThailand @XMRVoid yeah, pixel 8a is dirt cheap. graphene runs great on it.
English
@DanOpcode Yeah, I had mostly reached the same conclusion. There is a wrinkle for them here. At least some subset of consumers they've forced out make or influence purchasing decisions.
English
@uberboringguy Probably aiming at enterprises and governments.
English
Anthropic is burning their reputation in real-time. Nerfing subscriptions, poor performance of Opus 4.7, removing Opus 4.5 & 4.6, not allowing to use subscription outside Claude Code, accidentally leaking their vibecoded source code, testing removing Claude Code from $20 sub...
English
@ns123abc The whole thing is BS. The same behaviors are observable in third-party harnesses.
English
@jpschroeder @DarioAmodei @sama Nope. And, oh hey, hope you didn’t like that third-party harness you were using, either.
English
Ahem, @DarioAmodei and @sama - this is what honesty looks like.
I pay you $200/mo - at least tell me when I’m using a quant.
dax@thdxr
we've had a huge spike in OpenCode Go subscribers so we have people on the team entirely focused on securing more capacity we're growing faster than our providers are receiving GPUs we have options but might be a bit bumpy as we figure it out
English
@CodeAkram @ptremblay 100% this. Make the serving reliable and let me use a third-party harness. Instead, I canceled my max sub just like OP.
English
@ptremblay I was left without words when I cancelled my max 20x sub
It felt like I was gut punching myself
Again I will say this w/o remorse
Claude models are absolutely brilliant
The products around them which Anthropic ships are not :/
English
this is what happens when you screw up your customers. 🤷 looking at you Anthropic...
English
@Yuchenj_UW I was seeing this behavior in other harnesses (before they banned them, of course), so that rules out CC. Implies some level of bias to say so, but I don't think it's psychology, either. I've watched it change to single-line outputs/refusals mid-session in the past few weeks.
English
Seeing rumors that Claude Opus 4.6 got nerfed.
Usually this boils down to 3 cases:
- Unintentional. For example, a regression caused by changes in the inference stack or Claude Code. This is what evals are for before rolling out.
- Intentional “optimizations” (quantization, reduced reasoning). If so, say it. If users pay for a model, they should get that model.
- User psychology. The more you use a model, the dumber it feels.
English
@BenjaminHouy @opencode Pi is a fantastic harness, especially if you're working with multi-provider models (just not anthropic, at least without an api key 🤡). shittycodingagent.ai It is worth mentioning that if you rely heavily on MCPs, it might not be for you.
English
Going to give @opencode a try.
Don’t want to go full Codex and end up in another walled garden.
Any tips? Fav models?
Benjamin Houy@BenjaminHouy
Goodbye Claude. It has been a fun run. Call me when you actually work reliably.
English
@AlexEngineerAI Gemma4 is genuinely fantastic; and I do see its capacity to be on-par w/ Sonnet w/ sufficient ctx. My ? is what size ctx wdw are you running it at? Even on a Mac Mini with 64GB RAM, ~49k ctx wdw seems to be the ceiling (but I've still some optimization/testing work to do).
English
I cancelled my Claude subscription.
Gemma 4 is free, runs locally, and hits 80% ...
The gap is basically gone.
Why are you still paying? 💵💰
Alex Finn@AlexFinn
It’s over. Anthropic just banned OpenClaw. Uncensored thoughts: 1. Massive mistake that will come back to bite them 2. Open source needs to win. If you have a local model running on your Mac mini, no corporation will ever be able to ban you 3. ChatGPT 5.4 is the best model. But it sucks compared to opus in OpenClaw. I will continue to pay for Anthropic api 4. I have no doubt the next OpenAI model will be optimized for Openclaw and be excellent 5. In 6 months the local models will be as good as opus 4.6 and all of this will be forgotten 6. It’s feels like from a consumer sentiment perspective things have flipped for OpenAI and Anthropic. They were the darlings when Opus 4.5 came out 7. Going to the Kanye concert right now please don’t spoil the stage or set list in the replies 8. The best openclaw set up is now Opus as the orchestrator, then much cheaper models as the execution layer. If you do this properly you won’t be paying much more than $200 a month. I’m using Gemma 4 and Qwen 3.5 for execution on my DGX Spark and Mac Studio
English
@EffortDefines @HackingDave I will absolutely check it out. I have at least four public projects that were autonomously built outside the original spec authoring, so I feel like I've got a pretty good workflow established.
English
Codex spark is amazing tbh. Blows through usage but wow reminds you of how important speed is.
If you're looking to really dial in your workflow though, I'd try our agent orchestrator Animus (open source). It vastly improves performance once you've dialed in a workflow. Repo on my profile if interested.
English
Yeppers. It’s the worst model right now. When first released, incredible.
The fact that Claude hasn’t publicly stated this or what they are doing to fix it is not a good look. I’ve switched our dev teams away from Claude to Codex.
Chubby♨️@kimmonismus
AMD’s AI director Stella Laurenzo claims Anthropic’s Claude Code has significantly declined in quality since early March, citing analysis of 6,800+ sessions and 234k tool calls showing rising “laziness” behaviors like shallow reasoning, skipping code review, and incomplete tasks. Honestly, this is more impactful than expected, engineers report the model now favors quick, incorrect fixes over deep problem-solving, raising trust issues for complex workflows.
English
@aiHax_on_X @cgtwts Yeah, exactly, it's a compounding problem. Exactly what you said -- if they could've provided a consistent quality experience, I would've been ok with a smaller bucket, but I guess Anth had other plans...
English
@uberboringguy @cgtwts The combination of the strict limits and the degraded quality is what makes the main problem.
Claude makes mistakes / bugs -> reprompt -> even less limits.
With current changes i end up with more wasted time and less output
I rather want half the current limits but good output
English
Agreed. I guess I should've qualified "best models" with "when they're not 2-bit quantized". The overhead of constantly regrounding opus with the degradation lately was enough. But killing my oauth access with pi was what did it for me. I'm on
OAI now as well -- it's not bad!
English
@uberboringguy @cgtwts The models have been better.
With the current nerfs they are not anymore.
Claude makes so many mistakes that i can as well use openAI which on the other side has way more generous limits.
English
@jumperz There is definitely a place for a Cowork-like product in enterprises - for the non-dev type who wants automation/orchestration capabilities. And yeah, I cancelled my max sub b/c pi is a fantastic agent (and I can't use it on the "regular" sub bucket anymore).
English
pretty sure if codex had been more consistent in its voice and had a stronger identity, no one would be paying for claude.
im seeing so many people move on from claude no matter how many features they ship, the rate limits have been unbearable
and yeah, banning openclaw through the auth didn’t help either.
also, who’s actually been using cowork? and did it replace hermes or openclaw? no not even close
with all that being said, claude is clearly aiming for casual users and everything I just mentioned points in that direction so they rather sell more of $20 plans and if its not enough they push users to buy API tokens.. rather than selling their max plan.
I have no intentions of defending any billiondollar company but for real give me one good reason makes me “ need “ to use claude and obviously pay for $200 plan? especially that they also gatekeeping mythos too..
English
@Just1Conserv And anyone who’s being honest about this knows they tabled all the legislative tax hikes until next session (where they’ll pass without question) to avoid the bad midterms look. Looks like they decided to start with constitutional freedoms before coming for our cash.
English
"Spanberger hasn't raised any taxes ackshually"
She put us into RGGI, a tax on your power bill
She'll sign public employee collective bargaining, guaranteeing real estate, meals, and car tax hikes. Prob a sales tax hike before 29.
That's not even to mention business mandates.
English
@Griffin4VA @grok what law was signed by VA gov on 4/11/26?
English
