@rohit3a Haiku and Sonnet are really important strategic models, because those are the ones that will directly compete with open weight models. OpenAI isn’t the biggest threat to Anthropic, it’s all the Chinese companies.
@mstockton People don’t like change. AI has a bad rep in the media. Also, you’d be surprised how many technical people I speak to that are heavily resistant to AI. If we struggle to get all of our engineers to buy in, imagine how much more daunting it is to get the business roles in.
@omarsar0 Yeah not just easier to consume, it has that warm personality back. Personality is something that is hard to quantify, but it really has a big impact in how enjoyable the models are to interact with.
Opus 5 is the type of release that makes me wonder if we even need smarter models. We now have a cost efficient model that is a joy to interact with, and can handle anything I can throw at it. Maybe Dario is right and we should just pause here 🤣
@petergyang Opus 5.5 is the best model I’ve ever seen. It’s like they took my entire wishlist and built it all into a single release. Polar opposite of the 5 launch.
@IndependentEco It’s great! According to Anthropic’s docs medium is about equal to high in past models. So I’ve downgraded all of my workflows one effort level. They really killed it with this model on all accounts.
Watching all these people posting about the WoW Forever beta this weekend really makes me miss playing it back in the day. I start wondering if I should give it a go. Then by Tuesday I’m like “oh yeah, that’s right, I have 0 spare time.” Maybe one day when I retire 🤣
Yeah I think so. Also, I don't think they want people using Fable as a daily driver. I think they really want Fable to be reserved for the most complex high level problems only. So the reduced rate on Opus is helping to entice people to use Opus. Looks like Opus 5.5 is quite compute efficient. Even when we get a Fable upgrade, it's even harder to justify the price.
Opus 5.5 is a dream to work with. I was shadow routed for 2 days last week. It’s smart, token efficient, fixes the response issues from 5, and it’s cheaper. What’s not to love?
@tiwarisuhani_11 Most of those things are dead. But imo there’s seemingly infinite new problems to solve. So our jobs will become engineering solutions for those new problems. This is a very exciting and fun time to be an engineer. It’s a very frightening and poor time to be a developer.
It doesn’t make sense to add complexity until that complexity is needed. I feel like people oftentimes jump on new trendy things just because they are new and trendy. Every part of your architecture should have a legit reason to exist. Start simple. Scale out as needed, when needed.
Not sure where you’ve worked. I’ve been in EM roles for the last 10 years. Everywhere I’ve been I’ve alwya been held 100% accountable for everything everyone on my team does, every line of code, etc. If I promote someone to Staff it means they’ve really earned my trust and I’m willing to take the blame for their inevitable mistakes 🤣
@Siggis_Man@dizaytsev I’ve had some teams that were so huge (15 directs) that I had to just babysit. I hated it. I prefer smaller teams, like I have now. Where I can basically be involved in the high level architecture, research, planning, decisions, etc.
@JSRavencroft@dizaytsev I don’t know where you’re working but as a 10 year EM manager I’ve always been held accountable for everything anyone does on my team.
@LeoBuilds_ Yeah we have one still. But he’s been very forward thinking and is moving more towards kanban and away from SCRUM. Kanban does still work quite well.