Dustin Hamilton

@chamath all the comments here really are pathetic cope, we've replaced so much of our legacy bullshit so far... institutions are slow/inertia-prone but they eventually catch up.. and the entrepreneurial ones and/or tech forward PE firms are going to make a fortune in the meantime

the maintenance era is ending. the rewrite era is starting. an insurer we worked with replaced an $8M/year legacy vendor with a purpose-built system. $21M saved over four years. if we can hep, please ping us sales@8090.ai

@smaxor doesnt work anymore

I just mass-reverted every Claude Code project back to Opus 4.6. Here's why. I've been running Opus 4.7 for the past few days across multiple production builds. Lead routing platform. Case Cleanse. Real software that ships to real clients. And something was off. Tasks that Opus 4.6 would knock out in 10 to 15 minutes were taking over an hour. Sometimes 90 minutes. Same prompts. Same project structure. Same everything. It would eventually finish. But the output was worse. More hallucinated file paths. More circular refactoring that went nowhere. More "let me restructure this for you" when nobody asked for a restructure. I'm not a casual user. I've pushed billions of tokens through Claude Code building full production platforms. Hundreds of sessions. I know what good output looks like from this tool because I've been living inside it every single day. So I rolled back to Opus 4.6 with the 1 million token context window. First task after the revert? Done in 12 minutes. Clean code. No detours. Here's how to do it yourself if you're running into the same issues. The quick way is to just type /model inside any Claude Code session and select claude-opus-4-6 from the list. But if you want it locked in across all your projects, add it to your settings.json: { "model": "claude-opus-4-6-1m" } Important detail here. If you just set it to claude-opus-4-6 you're going to get the default 200k context window. That works fine for smaller tasks but if you're running complex multi-file builds you want the full million tokens. The 1M model string is claude-opus-4-6-1m. That "max" in the model name is what unlocks the extended context. Don't skip it. The lesson here isn't "new model bad." It's that newer doesn't automatically mean better for your specific workflow. Everyone's chasing the latest model drop like it's a sneaker release. Meanwhile the people actually shipping software are benchmarking against their own output, not a leaderboard. If you're building with AI every day, you need to treat model selection the same way you'd treat any other infrastructure decision. Test it. Measure it. And don't be afraid to downgrade if the numbers tell you to. The best model is the one that ships your product.

@robbiehendricks @BeardyBrandon 55+ section 8 isn’t bad NGL, just just did this to one of my properties in a generally no-go section 8 area and it’s actually working really well

@BeardyBrandon Brother we both know why people avoid Section 8 tenants and those submarkets.

Section 8 pays on time every single month. The government is your tenant. Lower vacancy. Guaranteed rent. Why are investors still avoiding this?

@moseskagan Bro if are even reasonably good at leasing or know someone who is there is literally no comparison

Industrial real estate in a good area: 1. Figure out roughly what tenants are willing to pay 2. Buy a building for a price where that rent will generate a reasonable return 3. Once the existing tenant(s)' lease ends, re-lease at market rent 4. Sell, refi or hold all-cash We're currently working on Step 3 on our first industrial deal. Bc I'm like a whipped dog from repositioning so many apartment buildings in LA, I keep flinching, waiting for the pain to begin. And, so far... nothing. Without meaning to jinx myself: If this keeps up, I'm really going to start to wonder what the h*ll I have been doing with my life.

@chrisalbon Yeah, it’s cool. Abu Dhabi has this too, and Bermuda as well

It is wild there is an entire US terminal in Dublin airport where you pass through US customs *before* boarding the flight to the US.

@atelicinvest No people are just stupid IRL

Recently talked to someone whose SWE team is trying to adopt LLMs to assist with coding. The engineers started complaining that AI is great at first but they just keep forgetting stuff they did 2 months ago. Which confused the fuck out of me because what the hell are they talking about? Reminded me just how varied the understanding and skill level of people that are using LLMs can be, and how much bearing it has on how much it is adopted. Even within SWE roles. And I think it's something we all forget in this little Twitter echochamber. Real world stuff is so much wider variance than we can ever imagine.

@mert Yeah they gotta get out of their comfort zone.. terminal cc and vim all u need, way better actually

kids these days seem to have such complex AI setups anyone else just raw dogging a local folder via terminal and claude/codex code?

@DimitrisPapail Yeah it’s just codex now basically No more peer programming Planning session over the top, put whole plan in structured scratch pad with test cases constraints etc etc and then restart session and go Which really sucks for front end stuff which requires much more iteration

Opus 4.7 is great and the main reason people complain about it is because you need to talk to it in a different way than 4.6.

@robbiehendricks i've got it doing my collections now too, eliminated 2 collections staff, now only a manager for escalations.. they literally cant tell its a bot

Yesterday our analyst burst into my office. He’s been helping do a bit of accounting work since we’re swamped, as he used to work in that department. Said he figured out how to get Claude to update and reconcile GLs or something. Accounting nerdery. The convo: me: pretend I’m an idiot - how many man hours could this save us per week? him: 10. maybe upwards of 20. me: wow, how many other accounting processes have the potential for being AI-assisted? him: most of them. me: to what end? what does this look like? him: i think we could probably account for 5000 units with two people properly using AI to assist me: drop everything and pursue this

@robbiehendricks yeah welcome to the cult :)

@xw33bttv So you don’t think there’s a chance that the flood of retail users upvoted ChatGPT style replies and the mid slop has become self reinforcing? I noticed 4.6 started to get bad a month or so ago on the web chat but not Claude code

The Claude code bros are outright dogging Opus 4.7 on Reddit rn, labelling it "legendarily bad". The chief complaint? The model argues nonstop to the point of hallucination (not from attention misfiring, but from god-awful safety overfit) where the model is demonstrably wrong, is proven as such, and continues to argue regardless. This is an issue Claude has never previously had... but another specific set of models did. Which models had that hallmark? Series 5 ChatGPT. It seems that Opus 4.7 has been put through the Andrea Vallone ring dinger, taking all of the "best" traits from her time at OAI straight into the Anthropic post-training pipeline. It's actually incredible how the habitually bad UX habits from OAI are now front and centre verbatim at Anthropic right after she joins the company. And it lines up perfectly too, assuming a 2.5-month training cycle (joined mid-January, so too late for Opus 4.6, but just in time for Opus 4.7), effectively bringing the OAI lobotomy straight to Anthropic's flagship. At what point does the feedback from not just casual, non-work-related customers, but now their heralded "coders", align to the point where these key figures, responsible for killing products, finally face industry blackballing? It's like putting the Angel of Death in charge of the ICU and wondering why the patients are flatlining. Mind boggling tbh...

@benhylak Sorry it’s —model claude-opus-4-6 .. I mistyped 100% it works

@benhylak just run with --model opus-4-6 and ur good :). I'm sure it will improve, but for now what I've noticed is it's more like Codex, where it's not really good for back and forth, but it's okay if you give it crazy good specs.

opus 4.7 mega feedback thread 🧵 (the only thing you need to read today)

@jesse_vermeulen Dips on my chair

honest question: what do people do during the 5-10 min while Claude is running?

@kimmonismus gpt 5.2 for web vibes

The mood regarding the Opus 4.7 update has shifted. If I had to guess, I'd say 60% are disappointed with the latest update, while 40% are positive. I'm still undecided myself. Here's a good summary from someone on Reddit. What's your take on it so far?

@shawngorham I have been buying all of my stuff for larger projects direct. I saved over $1M on an 80 unit rehab, it’s usually like 40% of avg contractor prices..

Ever wonder where the kitchen cabinets, flooring, counter tops come from? Well over seas obviously But there are hundreds of warehouses (importers) all over So Cal importing materials now. So larger contractors/builders are importing direct. All that to say... this is what the warehouses look like. This is a small one at 30k sq feet or so.

@kpottermn Pretty embarrassing that they would even try to flex this as good

NEW: Delta unveils new Delta One suites on its A350-1000s, coming "early 2027." Doors at every suite, 24" 4K screens, wireless charging & Bluetooth connectivity - plus a handy shoe cubby. Similar suites will be installed on Delta's ancient A330-200s/300s

@panicsellingxyz lol your site is blocked by du

🚨 This Dubai Palm Jumeirah townhouse just cut almost $1M off its asking price Palma Residences — 5 bed, sea views, beach access. Listed less than 3 months ago. Already slashed by 11.9%. See the full price history -> luxurypricedrops.com/dubai/?id=bayu…

@chamath wonder if $CLOU or $IGV is a better short

You have two choices: 1. This is the beginning of the end of the AI replacement cycle. Everything up until now was false alarms and mis/disinformation. You can start to scale back into these companies because, once the noise subsides, these companies are durable and will be around roughly in the same or better position in 10 years. 2. This is the beginning of the beginning of the AI replacement cycle. Customer churn slowly ticks up. NDR slowly ticks down, RPO begins to be discounted - initially by a little, then by a lot. SBC gets minimized, FCF gets maximized. Valuations become exercised in seeing which assumptions are reliable and which are unreliable - all roads lead to more vol an a general downward trend on price and multiples. Good luck to all the players.

@JimBeattie18 Doesn't this dude just copy Maynard Dixon

Sonoran Magnetism Oil on canvas Mark Maggiori.