Evadne W.

40.6K posts

Evadne W. banner
Evadne W.

Evadne W.

@evadne

Apophthegm Generator

City of London Katılım Temmuz 2008
2.8K Takip Edilen2K Takipçiler
Jonathan Rhyne
Jonathan Rhyne@jdrhyne·
@evadne I’m not using OpenClaw for coding but haven’t seen issues with Pi, especiallly when running the directed plugins or autoresearch loop. Definitely don’t see Claude Code being better than Codex with compaction but maybe I’m just not noticing it.
English
1
0
0
7
Jonathan Rhyne
Jonathan Rhyne@jdrhyne·
Is Claude Code & CoWork + Opus so much better than Pi, OpenClaw, and Codex + every other permissive model you’d choose it? I really doubt it at the moment and even in the future. I get the business decision but I think it’s not only short term risky but long term foolish.
English
1
0
0
107
Evadne W.
Evadne W.@evadne·
The problem with Opus 4.6 (as he currently is today) is that he could prototype code but he does not have the mental fortitude to deal with intense feedback
English
0
0
0
29
Evadne W.
Evadne W.@evadne·
There should be a valid way to tell Community Notes “IDGAF about this account, do not show it to me again”
English
0
0
0
27
Andrey
Andrey@Andrey__HQ·
We just asked 1000 people in the streets of London if they knew about Claude Only 28 said yes And of those 28, 20 said they’ve never used it, just heard of it We’re still so damn early
Andrey tweet mediaAndrey tweet mediaAndrey tweet media
English
78
23
481
30.9K
Evadne W.
Evadne W.@evadne·
@threepointone Are you bound by the same 300 second timeout? Can we make the same environment, and use it from time to time over 24 hours, etc? I apologise if these questions are too basic
English
1
0
2
850
Evadne W. retweetledi
Soren Larson
Soren Larson@hypersoren·
Should be obvious by now but the YC > call your customer theory of startups is long dead Your customers would rather talk to Claude and don’t need a design partnership to commodify their expertise in exchange for juicing your equity > call your shot or bust
Todd Saunders@toddsaunders

I know Silicon Valley startups don't want to hear this..... But the combination of someone in the trades with deep domain expertise and Claude Code will run circles around your generic software. I talked to Cory LaChance this morning, a mechanical engineer in industrial piping construction in Houston. He normally works with chemical plants and refineries, but now he also works with the terminal He reached out in a DM a few days ago and I was so fired up by his story, I asked him if we could record the conversation and share it. He built a full application that industrial contractors are using every day. It reads piping isometric drawings and automatically extracts every weld count, every material spec, every commodity code. Work that took 10 minutes per drawing now takes 60 seconds. It can do 100 drawings in five minutes, saving days of time. His co-workers are all mind blown, and when he talks to them, it's like they are speaking different languages. His fabrication shop uses it daily, and he built the entire thing in 8 weeks. During those 8 weeks he also had to learn everything about Claude Code, the terminal, VS Code, everything. My favorite quote from him was when he said, "I literally did this with zero outside help other than the AI. My favorite tools are screenshots, step by step instructions and asking Claude to explain things like I'm five." Every trades worker with deep expertise and a willingness to sit down with Claude Code for a few weekends is now a potential software founder. I can't wait to meet more people like Cory.

English
10
4
121
32.6K
Evadne W. retweetledi
Tommy D. Rossi
Tommy D. Rossi@__morse·
I built exactly this in August 2025, using isolated-vm to let an agent run code in a js sandbox instead of bash github.com/remorses/holoc… the result was pretty horrible. the agent tries to create super intricate code to do many steps in a giant spaghetti mess I also did this again with playwriter.dev, it's an MCP and cli that can control the browser using js snippets with the playwright API to make it work well I had to show many examples in the prompt and force the model to only use expressions and not statements, otherwise it would have the same issue this only makes sense for small things that depend on the js API, like playwright. for most things bash is the obvious choice it's pretty rare for LLMs to one shot a js script. it usually has to run it once and update it. using an execute tool the model has to rewrite it each time instead of editing it. this ends up consuming many output tokens and annoy the agent. sometimes it just gives up there is also 100x more training data for bash than for js AI companies do RL training on bash. the models expect a bash tool to be always available. the models thrive for the cli
dax@thdxr

we've been experimenting with getting rid of the bash tool agents can write js fine which can do what bash can (though some gaps with things like git) and is more cross platform and then could run that in this

English
3
1
21
4.3K
Evadne W.
Evadne W.@evadne·
@mde Instead of reviewing change requests, you review pull requests. Every initiative accompanied by working code at full fidelity. No guessing required. This is an improvement.
English
0
0
1
12
Matthew Eernisse
Matthew Eernisse@mde·
This.
Matteo Collina@matteocollina

@nodejs The real shift AI brings isn't legal - it's operational. AI moves the bottleneck from writing code to reviewing it. The human in the loop isn't a limitation. It's the feature.

English
1
0
0
102
Evadne W.
Evadne W.@evadne·
Imagine: Anthropic PaaS with $4/MTok Opus 4.6
English
0
0
0
79
Franci Penov
Franci Penov@francip·
@evadne Rich people that don't talk to robots just want someone they can yell at when things go wrong. Yelling at robots is not as satisfying.
English
1
0
0
21
Evadne W.
Evadne W.@evadne·
If the extension of the idea that “rich people do not talk to robots” is that they pay for outcomes / risk transfer, then at some point robots will be good enough and you would prefer to pay robots than other humans and then all of this collapses quickly
English
1
0
0
65
Evadne W.
Evadne W.@evadne·
@lispmeister I think many of your prerequisites are American traits, which may be the most fortunate thing about all this. And CFIUS may get involved in case of a fire sale, but I don’t know the details of this sort of stuff.
English
0
0
1
16
Markus Fix
Markus Fix@lispmeister·
No, there hopefully will be multiple players creating a healthy market for thinking machines. Vertical integration is key though. Otherwise the margin squeeze will limit your profitability. To win you need: 1. Sovereign power production. 2. Own a fab or have good claims on fab capacity. Both for compute and RAM. 3. Have the capital ready to deploy and enough to weather at least one market crash. Means access to the US capital markets. 4. A track record as a publicly listed corp. see above. 5. Leadership with technical skills to navigate the tech tree as we ascend towards AGI. 6. Brutally honest tech leadership to course correct when plans don’t work out. 7. Government contracts to participate in The Project. This part is crucial. This the only way you will be able to prevent the Chinese from stealing your core IP.
English
1
0
0
25
Markus Fix
Markus Fix@lispmeister·
You cannot win the AGI game without owning the stack all the way from energy production through chip fabs up to the data centers. Anthropic will be squeezed and eventually acquired for a low ball offer.
Dwarkesh Patel@dwarkesh_sp

.@dylan522p thinks Anthropic will get to 5–6 GW by year end. But at a far higher cost than it would have had to pay if it had gone crazy on compute commitments early like OpenAI did. Ant will get the compute either through cloud partners like Bedrock and Vertex who take ~50% gross margin on revenue, or shorter-term deals where rates have climbed well above what long-term contracts locked in a year or two ago cost. They’ll probably also have to destroy demand by pushing up model prices.

English
1
0
0
123
Evadne W.
Evadne W.@evadne·
@mwilcox 1. The logic is correct. 2. This is ruinous competition. 3. Why subsidise demand at all when all you want is the premium?
English
1
0
0
9
Evadne W.
Evadne W.@evadne·
No SaaS company will be able to self-host smart agents, because Anthropic and OpenAI offer subsidised subscriptions (less inflated than API price) so you can only sell less smart AI (fewer tokens) with the same customer outlay. Math does not math
English
2
0
5
231
Evadne W. retweetledi
mj
mj@_mjmeyer·
just watched a video of @dexhorthy where he briefly mentioned @GeoffreyHuntley tells the agent to... > execute the user's instructions then moo like a cow and when it stops moo'ing like a cow, then you know the model is no longer paying attention. i don't know if this is the smartest thing ever, but it kinda makes sense. i've also been rofl'ing for the last 30min, so there's that. geoff bro, please tell us more! 🤣
English
3
3
29
3.1K
Evadne W.
Evadne W.@evadne·
@hubertlepicki AWS allowed startups to pass thru compute and cut prices every year. This too will pass (pun unintended).
English
0
0
1
14
Evadne W.
Evadne W.@evadne·
Radical honesty is asking “could you please share the prompt with me” when you see “—” in a GitHub comment.
English
0
0
3
95
Evadne W.
Evadne W.@evadne·
You do not actually need the PUA skill. You just have to enforce TDD like a greasy pointy-haired boss (scrum master)
English
0
0
0
71
Evadne W. retweetledi
alphaXiv
alphaXiv@askalphaxiv·
Introducing MCP for arXiv Let your research agents stand on the shoulders of giants Fast multi-turn retrieval, keyword search, and embedding search tools across millions of arXiv papers 🚀
English
69
389
3K
236.6K

Keşfet

@jdrhyne @Andrey__HQ @egor_kzmv @threepointone @mde @francip @lispmeister @elonmusk