News for hackers. Stuff that matters.
Re:Ask HN: What's your current AI agentic setup? by Animats on Thursday November 05, 2026 @03:39AM (#50376720)
Embarrassingly, I don't have one yet. I'm mostly still writing code one keystroke at a time. I've used Github Copilot to translate between languages and look things up, but haven't trusted it to write new code or fix anything.

Claude Fable is now good enough that I should be using it. The latest Claude version is said to be comparable to a mid-career graphics programmer, which is about what I need. Six months ago it wasn't good enough to use for such work, and probably would have made a hash of my code. There's been progress.

But Anthropic's business practices are scary. They have a reputation for cancelling accounts for random reasons, offer rather bad terms, and have no customer support. I have a strong aversion to services with a reputation for cancelling services arbitrarily. They will apparently cancel based on their system's guesses as to what you're working on. Sometimes they guess wrong. What I'm doing (metaverse clients) is so far out of the mainstream they may guess wrong.

I'm currently stalled signing up for Claude because their verification email system can't get to my inbox. (Some problem with their using Constant Contact, a spam system, to send transactional emails to my own domain, which, long ago, I had blocked at the Constant Contact end. I think. The verification email never touches the inbound mail server.) It's very important to Anthropic that you sign up with the "right" email address, and it may not be possible to change it, so using some disposable address is not a good idea.

Suggestions?

Re:Ask HN: What's your current AI agentic setup? by olmo23 on Thursday November 05, 2026 @03:54AM (#50376839)
Just use a random gmail address, it works.
Re:Ask HN: What's your current AI agentic setup? by Forgeties79 on Thursday November 05, 2026 @09:53AM (#50379822)
If the concern is business practices, privacy, etc. adding Google to the mix doesn’t seem great
Re:Ask HN: What's your current AI agentic setup? by fidelramos on Thursday November 05, 2026 @06:39PM (#50386164)
Or use SimpleLogin.
Re:Ask HN: What's your current AI agentic setup? by Animats on Saturday November 07, 2026 @12:52AM (#50402024)
Claude account creation turned out to work with a domain I have hosted on Dreamhost. Although Dreamhost works, emails to my domains hosted on shared hosting at Hostgator are never replied to by their verification system.

Anthropic's support chatbot is happy to talk to me using a Hostgator email address, though. It says it doesn't have access to block information.

Somewhere Anthropic probably has a blacklist related to something on Hostgator shared hosting. Maybe I'll get my sites there moved to a different server.

Re:Ask HN: What's your current AI agentic setup? by walrus01 on Thursday November 05, 2026 @04:19AM (#50376993)
I would suggest that based on your current level of "trust" in these things (justifiably very low, I use zero openai or anthropic products myself) that you look into self hosting locally something reasonably capable and smart on your own hardware, on llama-server or something functionally equivalent to it. And running it in something like the opencode harness, using zero of the opencode API or online services. Qwen 3.8-flash-next will run on a lot of different hardware setups that are within the reach of normal people, or possibly deepseek v4 flash 1017, glm5.3-flash, things in that general category.
Re:Ask HN: What's your current AI agentic setup? by sampullman on Thursday November 05, 2026 @08:03AM (#50378605)
Qwen 3.8-flash-next is great, but if they say SOTA models from 6 months ago weren't good enough for the kind of code they write, Qwen isn't going to cut it yet.
Re:Ask HN: What's your current AI agentic setup? by Animats on Thursday November 05, 2026 @03:25PM (#50384003)
I'm not doing enough to rent my own GPU farm.
Re:Ask HN: What's your current AI agentic setup? by computomatic on Thursday November 05, 2026 @04:36AM (#50377140)
You can access Claude models through any number of providers - OpenRouter, AWS Bedrock, etc. and as far as I know, Claude Code can be used with any of them. The only real benefit to going direct through Anthropic is the subscriptions. It’s worth the risk for the discount, but if you get locked out, there’s other (pricy) options to get unblocked.
Re:Ask HN: What's your current AI agentic setup? by ChrisRR on Thursday November 05, 2026 @05:14AM (#50377407)
You don't even need to use Fable. Give Opus 5.5 a try
Re:Ask HN: What's your current AI agentic setup? by nedt on Thursday November 05, 2026 @08:05AM (#50378628)
I have a longer history of agents sessions and used those to benchmark the different models. Sonnet 5.5 is already pretty strong, but not always as careful (leaked secrets into the session). Opus 5.5 is already able to solve everything I'm doing. You might not need Fable for anything. Biggest different between Opus and Fable in my benchmarks was price. Also Opus 5.5 is cheaper now then previous Opus so you might also not need Sonnet as much for cost savings.
Re:Ask HN: What's your current AI agentic setup? by matltc on Monday November 09, 2026 @05:32PM (#50433229)
Is your benchmark open? Can I get a link if so?

If not, did you parse the .jsonl files in the projects/ directory?

Re:Ask HN: What's your current AI agentic setup? by nedt on Monday November 09, 2026 @07:02PM (#50434217)
Did I parse the jsonl files? No. I just asked Claude to create a skill to read all previous conversations across all projects or anywhere else in Claude, Cursor, Github Copilot and Windsurf. For the last one it failed (they are encrypted). The rest it just picked up whatever location and format they are stored in. So yes the skill is parsing the jsonl files in projects/. Only exception are Claude Desktop sessions that are only stored in the cloud.

The skill itself is also the basis for my memory across sessions and projects. With that Claude can recall what I can remember. It's using a subagent so only the important bits land in context.

From there it's easy to get some benchmarks. Let it find common themes across your sessions and let it mock them to create benchmarks strong enough so not everything gets the same score. My benchmarks are now based on me designing new tools, debugging live issues going through log files, finding clues in bigger code bases etc. pp.

But no they are not open. First of all they depend on my use cases (but I also got one colleague to contribute cases) and second of all they are based on the work at my company, so I can't share them.

Re:Ask HN: What's your current AI agentic setup? by nycdatasci on Thursday November 05, 2026 @09:23AM (#50379420)
Based on blended benchmarks from Artificial Analysis, Opus 5.5 is stronger than Fable 5.1. Are you finding use cases where Fable is superior?
Re:Ask HN: What's your current AI agentic setup? by felixgallo on Thursday November 05, 2026 @10:07AM (#50380023)
Fable can go deeper than Opus in my experience. Opus is incredible for relatively now-simple things like coding, but Fable has delivered improved results for me in data analysis on relatively obscure topics. That said one gets much more Opus for a subscription than one does Fable, so obviously benchmark your own cases and come up with your own cost benefit analysis.
Re:Ask HN: What's your current AI agentic setup? by Conez on Saturday November 07, 2026 @10:55PM (#50410300)
100%
Re:Ask HN: What's your current AI agentic setup? by mlikelihood on Tuesday November 10, 2026 @05:13PM (#50447027)
this. Fable is an overkill in most cases given its high price. It's better, just not as better for its cost.

Have been using Codex + Claude (their $100 plans) and for some time thought Claude was too slow and wordy so using Sol for day to day; then Opus 5.5 turns out to be still the best in quality and they've caught up in speed recently as well.

I think it's always best to have multiple agents review or discuss on the problem then delegate to a faster/lower cost model for execution.

Re:Ask HN: What's your current AI agentic setup? by creativeSlumber on Thursday November 05, 2026 @07:35AM (#50378366)
> But Anthropic's business practices are scary. They have a reputation for cancelling accounts for random reasons, offer rather bad terms, and have no customer support. I have a strong aversion to services with a reputation for cancelling services arbitrarily. They will apparently cancel based on their system's guesses as to what you're working on. Sometimes they guess wrong. What I'm doing (metaverse clients) is so far out of the mainstream they may guess wrong.

Then use open alternatives as much as possible. Something like Opencode with OpenRouter with Deepseek 4.1 from ZDR provider works very well and costs like 1/20th.

Re:Ask HN: What's your current AI agentic setup? by scosman on Thursday November 05, 2026 @10:18AM (#50380185)
It costs 1/20th the price at token pricing, but an Ultra subscription is still amazing value.
Re:Ask HN: What's your current AI agentic setup? by jpardoe on Thursday November 05, 2026 @10:32AM (#50380381)
In practice all of these open models are more expensive than a Claude subscription.
Re:Ask HN: What's your current AI agentic setup? by ggregoire on Thursday November 05, 2026 @11:27AM (#50381117)
Are they? (genuine question, i've got no experience with Claude) This month I used 1 billion tokens (including cache) with DS4.1 flash and it cost me 7% of my $10 opencode go sub.
Re:Ask HN: What's your current AI agentic setup? by jeffyaw on Thursday November 05, 2026 @03:33PM (#50384114)
i found they were until recently. mimo 2.6 flash and glm 5.3 flash seem to finally be cheaper. i find ds4.1 flash still on the high side.
Re:Ask HN: What's your current AI agentic setup? by jpardoe on Friday November 06, 2026 @09:21AM (#50392594)
I found GLM Flash competitive price wise but then my tasks took twice as long to complete. Even if it was cheaper (in my experience it still wasn’t) you then lose out on speed because of the insane amounts of inefficient reasoning.
Re:Ask HN: What's your current AI agentic setup? by jpardoe on Friday November 06, 2026 @09:19AM (#50392572)
In my experience yes. Been using Claude since day one and wanted to experiment with Pi for fun, and try to get my costs down. I found the open models more expensive, eventually had to plug my Pi set up into an OpenAI subscription. That was the only way to get good value… Then I realised OpenAI is way, way behind Opus 5.5 so I’ve come full circle and am back on Claude Code.
Re:Ask HN: What's your current AI agentic setup? by bloppe on Thursday November 05, 2026 @04:27PM (#50384691)
I also think OAI is much better than Anthropic in this regard. They don't try to gatekeep nearly as much.

But I agree that open alternatives are sort of a theoretical ideal. Just a lot harder to get those mega-subsidized.

Re:Ask HN: What's your current AI agentic setup? by Fervicus on Thursday November 05, 2026 @12:25PM (#50381788)
It's crazy to me that this is considered embarrassing now. People should stay up to date with current tech, but we shouldn't be ashamed of not using LLMs. Let's re-normalize coding by hand.
Re:Ask HN: What's your current AI agentic setup? by Animats on Thursday November 05, 2026 @04:03PM (#50384431)
I did, for a long time, but we're past that point now. Someone else got something done in three months that took me years of hand coding.
Re:Ask HN: What's your current AI agentic setup? by bloppe on Thursday November 05, 2026 @04:32PM (#50384741)
If something took 3 months, then there must have been a good deal of manual attention involved, even if it wasn't typing each character by hand.

I still find that it's often faster to combine manual coding with generation. I often start by setting up a skeleton with a bunch of TODOs, then tell my agent to fix each TODO. If my agents can't seem to stop coming up with insanely massive diffs for what should be small changes, it's often much faster to just code it properly by hand.

Re:Ask HN: What's your current AI agentic setup? by steve1977 on Thursday November 05, 2026 @04:36PM (#50384791)
And thinking by brain.
Re:Ask HN: What's your current AI agentic setup? by martin82 on Thursday November 05, 2026 @09:02PM (#50387225)
Coding by hand, sure! But let the LLM at least auto-complete your thoughts. They are still your thoughts, but it's just 100x faster than coding by hand.
Re:Ask HN: What's your current AI agentic setup? by 946789987649 on Friday November 06, 2026 @03:34AM (#50389676)
I was an AI-skeptic for a while but it is embarrassing to not even try it. I'm still on the more manual end of AI use. My prompts are very small, targeted changes. I still review every line as though a human wrote it. I know exactly what I want and I've got my prompts and claude.MD at a point where it gives me that most of the time. I probably only write 5% of the code now.
Re:Ask HN: What's your current AI agentic setup? by mirmor23 on Friday November 06, 2026 @12:32PM (#50395473)
> But Anthropic's business practices are scary. They have a reputation for cancelling accounts for random reasons, offer rather bad terms, and have no customer support. I have a strong aversion to services with a reputation for cancelling services arbitrarily. They will apparently cancel based on their system's guesses as to what you're working on. Sometimes they guess wrong. What I'm doing (metaverse clients) is so far out of the mainstream they may guess wrong.

From your response it does not seem like you have enough experience using LLM, and your response seem to indicate a bias from reading posts of folks who got banned. Typically in these situations the poster who got banned, will never typically reveal the full picture of what happened or what their setup was.

Anthropic customer service is terrible, their policies are opaque and could be better. But it is really delusional to think that they are kind of fly-by-night all-you-can-eat buffet that kick you out for no reason or just because you used up your quota.

From what I understand, if you are using official harness, not using it to build/offer competing service, not sharing account with others, and not doing illegal activity (this is left to their interpretation), the chances of being banned is quite low. This is not anything special to Anthropic, most service providers have some kind of ToS.