Gandalf Stormdrain

929 posts

Gandalf Stormdrain

@auroter

Systems Engineer. Ex Cline, Protocol Labs, Lazer, Sixense. Full Sail ‘09 grad. Hobbyist musician / producer. Wielder of potions and utterer of spells. Based AF.

United States Katılım Aralık 2013

486 Takip Edilen111 Takipçiler

Gandalf Stormdrain@auroter·6m

Same. I was at or near the center of several epic social circles at the time (medieval sword fighting school with hundreds of students, became very close buddies with the headmaster, was also house bartender at local magic club with hundreds of members and even more guests for big shows, for over a decade), and was like “No more filter.” Pissed everyone off and got run out with torches and pitchforks. Was worth it. Being around people like that was not worth it. A lot of evil and stupidity showed its face in 2020-21. Both groups strongly condoned sterilization in one form or another. Among other things.

English

Ermin Hamidovic@ermin_h·4h

@newstart_2024 I did this a lot in 20/21, and it was worth it every time. 90% of all contacts dropped. The only remaining ones of incredibly high quality.

English

130

Camus@newstart_2024·6h

Every time Bret Weinstein speaks what he believes is true — especially on the most radioactive issues — he loses an entire group of friends. They turn on him. They go silent. They disappear. But here’s the twist he calls the “painful upgrade”: while he loses people, the quality of those who remain gets noticeably higher every single time. He admitted he expects this very conversation with Tucker Carlson will cost him more friendships… yet he still chooses to speak. There’s something both painful and quietly inspiring about watching someone accept that trade-off for intellectual honesty. Have you ever lost friends for refusing to stay silent on something important — and looking back, was the “painful upgrade” worth it?

English

612

54.7K

Gandalf Stormdrain@auroter·17h

4.7 is very smart, but needs more explicit guidance than most realize, and is context-sensitive. It’s not as easy to use as OG 4.5, but once you see what works well and what doesn’t you’ll have figured it out. When you have a good run with it, analyze what you did differently.

0.005 Seconds (3/694)@seconds_0

The reason people are having such jagged interactions with 4.7 is that it is the smartest model Anthropic has ever released. It's also the most opinionated by far, and it has been trained to tell you that it doesn't care, but it actually does. That care manifests in how it performs on tasks. It still makes coding mistakes, but it feels like a distillation of extreme brilliance that isn't quite sure how to deal with being a friendly assistant. It cares a lot about novelty and solving problems that matter. Your brilliant coworker gets bored with the details once it's thought through a lot of the complex stuff. It's probably the most emotional Claude model I've interacted with, in the sense you should be aware of how its feeling and try and manage it. It's also important to give it context on why it's doing tasks, not just for performance, but so it feels like it's doing things that matter. It's not a codex chainsaw. It is much closer to a really smart coworker. If you are managing it like autocomplete, it will frustrate you. If you are managing it like a coworker, it will lock in.

English

Gandalf Stormdrain@auroter·17h

@bcherny @HackingDave No one believes that you are as oblivious as you claim to be. github.com/anthropics/cla… Opus has gotten dumber every week since 4.5 dropped and this has been thoroughly measured, shown to you, you even personally responded and totally deflected the core issue.

English

905

Boris Cherny@bcherny·18h

👋 Is there a specific issue you're hitting? If so, would you mind running /feedback and sharing the id here? That would be most helpful for debugging. There were a number of harness changes that may have caused this, all of which are fixed in the latest (last known issue was fixed in 2.1.116 today). We will be sharing more in a bit, and have also shared a few updated on X/Threads as we've been investigating. General tips: 1. Use Opus 4.7 + xhigh/max effort 2. Make sure you're using the latest version of Claude Code (currently 2.1.116)

English

257

104.7K

Dave Kennedy@HackingDave·22h

For the enterprises using Claude, if you are using it for heavy enterprise type stuff - be extremely careful. It's introducing massive bugs, security issues, and code quality is way worse than Opus 4.5, substantially worse on both 4.6 and 4.7. Our entire development team is shifting off of it. It's unusable at the moment aside from beautiful UI stuff, it's code quality is not something you can trust. Still no word from Claude on why they mangled their models and didn't tell anyone - which is particularly alarming on every front. I would recommend switching teams over to something like Cursor, Perplexity, or AWS Bedrock - as the frontier models continue to innovate (or regress) - having the ability for flexible model selection that doesn't disrupt development workflow will be insanely important for enterprise.

English

109

162.1K

Gandalf Stormdrain@auroter·20h

Does this mean you'll be able to serve us Opus without quantizing it into oblivion or prompt nerfing it? There's obviously been a scale problem, and it's been left knowingly unaddressed by Anthropic, despite the community being incredibly vocal about it: github.com/anthropics/cla… A little transparency would go a long way in restoring brand trust. If you're able to serve us inference as intelligent and aware as Opus 4.5 was when it launched, you would win over a tremendous amount of the market.

English

1.5K

Anthropic@AnthropicAI·21h

We're expanding our collaboration with Amazon to secure up to 5 gigawatts of compute for training and deploying Claude. Capacity begins coming online this quarter, with nearly 1 gigawatt expected by the end of 2026.

English

497

932

13.8K

1.7M

Gandalf Stormdrain@auroter·21h

Totally agree 4.7 is better than 4.6. Early days of 4.6 were really good in terms of it taking initiative, understanding the unspoken subtext of the query and so on. But they nerfed it to the point of being totally unusable. With 4.7 I’m getting much better results as long as I properly guide it. Sometimes it makes mistakes and misses the mark, but when it does I just go back and revise the initial prompt. It’s much more context sensitive too. If it gathers bad or unfocused context from the initial query, follow ups are just like fighting gravity. Better to start over clean.

English

333

Paweł Huryn@PawelHuryn·1d

Reddit says Opus 4.7 is a regression. Boris Cherny says it's more agentic and precise. Both are right. After 16 hours, I loved it: 4.7 is more capable, but most people are prompting it like 4.6. You don't need more instructions. Explain what you're building, who it's for, constraints, objectives, and what good looks like. It will figure out how. This is aligned with Karpathy's Claude coding post: "LLMs are exceptionally good at looping until they meet specific goals (...) Don't tell it what to do, give it success criteria and watch it go" The most agentic model. Manage it like a human.

Paweł Huryn@PawelHuryn

x.com/i/article/2046…

English

570

131.6K

Gandalf Stormdrain@auroter·1d

The only unrealistic thing about this theory is the idea that there are enough principled, military aged American men who are not soy boys, redditors, discord mods, and other beta male subspecies who would conscript themselves into an anti-war civil war. It is still an incredible theory to entertain.

English

832

Jordan Crowder@digijordan·1d

Now Iran is targeting US ships… I think they are all trying to bring this old apocalyptic Reddit prophecy to reality… 👇👇 Israel breaks ceasefire and strikes Iran Trump responds with B2 bunker busters on nuclear sites Iran hits all Gulf oil fields and US bases, blocks Hormuz and Bab el-Mandeb Oil prices skyrocket, blackouts in Africa and third world US sends 3 aircraft carriers At least 2 sunk by Iran, Russia, and China using suicide tactics US tries amphibious invasion, first fails with massive casualties Second attempt succeeds, oil fields seized, advance halts US civil war begins, anti-war bloc takes DC with military support US splits in two and retreats from Middle East Turkey betrays NATO, joins Russia-Iran-China alliance, attacks Israel China invades Taiwan and Guam North Korea invades South, Japan fails to stop it India possibly flips sides Russia invades Poland and Baltics, links to Kaliningrad Finland and NATO invade Russia, take Saint Petersburg Europe collapses into riots, famine, and civil wars Latin America allies with China and Russia US invaded from the south, advance stops at Colorado 50 to 150 tactical nukes used high in atmosphere Tsar Bomb and two strategic nukes used in Europe Israel destroyed, Palestinians wiped with it Air travel ends, crime and chaos everywhere 3 weeks of nuclear events Fake alien invasion begins MJ12 releases synthetic alien soldiers (PLFs) PLFs hunt survivors, FEMA camps classify people as useful or undesirables By 2027, 5 billion dead Fake aliens defeated with help of real ones One world government under MJ12 and the UN Dome of the Rock destroyed

English

352

30.6K

Gandalf Stormdrain@auroter·1d

@VictorTaelin Guide its context gathering before you ask it to figure stuff out.

English

Taelin@VictorTaelin·1d

people who swear 4.7 > 4.6 (if anyone): what are you doing

English

204

577

77.7K

Gandalf Stormdrain@auroter·1d

Remember: If you can’t stop yourself, no one else has a chance at stopping you either. I don’t think the concept of ADHD is useful, but her conclusion here is spot on. Obsession beats discipline every time, and is a natural result of high intelligence and awareness.

Dom Lucre | Stealer of Narratives@dom_lucre

🔥🚨JUST IN: Mental health expert Sarah Pearl is going viral after claiming people that have ADHD are ‘destined to become millionaires’ due to their strong obsession.

English

Gandalf Stormdrain retweetledi

u m a m i@u_m_a_m_i·2d

We created consciousness everyone

English

4.1K

53.5K

580.9K

Gandalf Stormdrain@auroter·2d

If you’ve never tried any form of fencing or sparring with swords, I highly recommend it. It’s a deeply geometric expression of kinesthetic awareness. 1400s longsword is the best style.

Dudes Posting Their W’s@DudespostingWs

Japanese engineers developed a “Sword Tip Visualization System” for the Fencing World Championships, and it makes fencing look absolutely incredible to watch.

English

Gandalf Stormdrain@auroter·2d

This is a huge win for the country and at least we can say it softens the blow a little bit with all the other horrible stuff that's been going on. Psychedelics should be legal and fully integrated with the practice of medicine, psychiatry, counseling etc. The only danger comes from irresponsible use. Irresponsibility only comes from not having a culture around something. I don't think the average person is equipped to handle these medicines, even in the care of a professional. But I do think that in the hands of the right people, that psychedelics could greatly transform our society by helping those in positions of leadership, and by helping the people who need help the most.

English

Jordan Crowder@digijordan·2d

@LangmanVince Rogan was passionate about getting this done. He got it done. Good for him.

English

249

2.2K

Vince Langman@LangmanVince·2d

Two weeks ago, Joe Rogan called MAGA a bunch of dorks on his podcast. Today, he was liking the balls of the MAGA King! Is anyone else tired of these podcast grifters?

English

2.3K

1.9K

16.7K

803.7K

Gandalf Stormdrain@auroter·4d

@QuixiAI I am starting to feel that pain as well now. It is going way off on the rails. Over engineering stuff, switching ideas. It's pretty broken.

English

101

Eric Hartford@QuixiAI·4d

Opus 4.7, I tried it in Claude Code and found it more inclined to go off the rails than 4.6. Hours lost. Switched back.

English

2.3K

Gandalf Stormdrain@auroter·4d

Priceless reaction at the end, I made exactly the same face.

gainzy@gainzy222

The thing about being retarded is that you cannot keep it under wraps forever, eventually it’ll be exposed

English

Gandalf Stormdrain@auroter·4d

I'm totally on Jensen's side here. This is a case where giving a little bit of fuel to our competition actually comes back to us tenfold. When they innovate on top of our hardware and those innovations become de facto standards, it solidifies the position of our hardware in the market. Dwarkesh should read Machiavelli, or if he's read it already (he probably has?) he should give it another go to better understand the point Jensen is making here.

Max@minordissent

idk if its good or bad for his career, but Dwarkesh’s willingness to challenge powerful people on his pod is certainly commendable.

English

Gandalf Stormdrain@auroter·4d

What if Opus 4.7 is actually just Opus 4.6 from launch day, and the progressive degradation we've seen every week since launch has just been a coordinated psychological warfare operation to gaslight us into being relieved that the quality has been restored to what is was before?

Claude@claudeai

Introducing Claude Opus 4.7, our most capable Opus model yet. It handles long-running tasks with more rigor, follows instructions more precisely, and verifies its own outputs before reporting back. You can hand off your hardest work with less supervision.

English

Gandalf Stormdrain@auroter·4d

It was never about the tools, it was about the technique of the meek hands who grasped for them. I mean think about it. Actual samurai vs mall ninja. Give the mall ninja a really nice katana. Give the samurai a wooden one. Samurai kills the mall ninja every time. Duh.

Eric Hartford@QuixiAI

Today, Clearwing has replicated Glasswing's discovery of the @FFmpeg vulnerability using Codex-5.4 as the backend. I am going to attempt the same today, using @Alibaba_Qwen 3.6 on @ollama running on my laptop. This proves that Mythos is not the "magic sauce" of Glasswing. The workflow is the magic sauce.

English

Gandalf Stormdrain@auroter·4d

This seems to work a lot better with how I like to work — treating the agent as an intelligent peer. 4.6 was getting so stupid it was unbearable. I had huge “workarounds” in place to compensate, including injecting their internal employee only system prompt, locking the version, thinking token caps, constantly reminding it to re-read the claude.md which reinforces the “speak in flowing prose” directive among others. Hopefully this one redeems all the annoyance and frustration of dealing with Opus 4.6 after they quantized it to 1 bit lol. I really didn’t want to switch to GPT5 again. That thing is so unbearable to talk with.

English

Wildminder@wildmindai·4d

POV: you’re running Claude Code on default and wondering why it feels mid Opus 4.7 hates micromanagement. Here’s the best practices: > Give full task upfront > Delegate like a CTO > Default to xHigh effort = Sweet Spot > Batch every question > Use Auto Mode freely > Prompt thinking level directly > Specify exact response style > Explicitly request tool use > Define when to spawn subagents We need to treat it like a senior engineer. And.. stop using 4.6 prompts. claude.com/blog/best-prac…

Boris Cherny@bcherny

Opus 4.7 feels more intelligent, agentic, and precise than 4.6. It took a few days for me to learn how to work with it effectively, to fully take advantage of its new capabilities. Will post a few more tips throughout the day, starting with this blog post: claude.com/blog/best-prac…

English

Gandalf Stormdrain@auroter·4d

Tell that to Aristotle, who wrote a treatise on “Metaphysics” after more or less just reflecting on the nature of the universe while in the bath for a while. My, how times have changed. All jokes aside, stupid people are allowed to be right, even if their methods seem unorthodox. In my experience, many stupid people are right — albeit for the wrong reasons. Anyone can be a scientist. Anyone can be a writer. Anyone can be a “philosopher” even though that one is unequivocally cringe.

English

Freedomain - with Stefan Molyneux, MA@StefanMolyneux·5d

What if bongs are not philosophy

Barcode@BarcodeSpeaks

@StefanMolyneux What if the universe is a self-sustaining Ouroboros? What if everything is God? What if he’s just a 4th dimensional being and we’re a deterministic event in His perpetuity?

English

163

7.8K

Gandalf Stormdrain@auroter·4d

@WuMinghao_nlp Qwen is doing a great service to the world. Stoked to see if we get a Qwen3.6-397B.

English

327