Beer van der Drift

116 posts

Beer van der Drift banner
Beer van der Drift

Beer van der Drift

@BeerDrift

Building things through software

Berlin Katılım Nisan 2020
87 Takip Edilen45 Takipçiler
Beer van der Drift
Beer van der Drift@BeerDrift·
I successfully trained my first model! Wouldn't dare call it a large language model er even a small one. I dub this one Timmy; a Teeny Language Model. Timmy only knows 🌈 rainbow colors and training him has been a blast. On top of that the biggest surprise was how similar Timmy learned like a human; simple stuff first and hard stuff later! This was purely an exercise for me to understand better how this process works under the hood. ✍️ More technical write-up in this article: x.com/BeerDrift/stat…
Beer van der Drift tweet media
English
0
0
1
15
Beer van der Drift
Beer van der Drift@BeerDrift·
Imagine the amount of unnecessary panic because of todays Tanstack hack. npm decided to set all versions matching '>= 0' as hacked leading everyone everywhere to get a positive on running npm audit. I spent hours rotating production secrets and notifying customers on a secondary secure laptop while it turns out I wasn't impacted at all.
Beer van der Drift tweet media
English
2
1
2
347
Beer van der Drift
Beer van der Drift@BeerDrift·
@andre_quentin I suspect this is something on the hotel's side. I had this exact same phishing attempt happen after booking directly through a hotel. The phishing site was a fake Agoda site even though I had booked directly. Very scary though to see how many details they already had.
English
1
0
57
15.4K
Quentin André
Quentin André@andre_quentin·
Things are about to get real scary. I got a WhatsApp message about a hotel reservation I made on Booking.com. They have my correct hotel name, correct dates, correct amount, everything. They ask me to confirm my CC details, without which they'll cancel my res.
Quentin André tweet mediaQuentin André tweet mediaQuentin André tweet media
English
362
372
3.6K
1.9M
Beer van der Drift
Beer van der Drift@BeerDrift·
@venturetwins Curious to hear your definition of AGI here. How I view it true AGI should excel like a human given it would only have access to a CLI. None of the current models are there yet.
English
0
0
0
17
Justine Moore
Justine Moore@venturetwins·
I suspect we already have AGI with the current models for many use cases - but the harnesses just aren’t there yet
English
126
39
793
103K
Beer van der Drift
Beer van der Drift@BeerDrift·
The metaphor I use is building with duct tape vs concrete. Creating a tipi (building prototype) with duct-tape is fine! Or adding a balcony w/ ducttape to a proper building you'll also get a happy result. But trying to build a skyscraper with ducttape will at one point fail. You need to consider the boring stuff; elevators, waste management, airconditioning. As AI gets better this threshold moves. Maybe it won't be duct tape anymore, but ABS plastic, and one day it'll be concrete-equivalent. But today we're not there yet.
English
0
0
0
56
Mariusz Kurman
Mariusz Kurman@mkurman88·
After two months of heavy "coding" with AI agents, I have one conclusion: if your codebase already exists, is fully human-written, and you use agents to add or improve features, it works great. However, when you try to create something new from scratch, they tend to add so much overcomplicated spaghetti code that it's hard to maintain in the long run. No matter which coding model you use, sooner or later, you'll hit a wall you can't break through.
English
502
376
5.6K
477.9K
Teri Beckham
Teri Beckham@TeriDBeckham·
@BeerDrift Beer van der Drift Wow! Coolest name on the internet.
English
1
0
1
25
Beer van der Drift
Beer van der Drift@BeerDrift·
@tpschmidt_ Did you run into any issues around your source IP now coming from known datacenters triggering more bot filters?
English
0
0
0
530
Tobias Schmidt
Tobias Schmidt@tpschmidt_·
I stopped running OpenClaw locally and moved it to a $5 Lightsail instance on AWS. Works surprisinlgy well. No need to run anything at home and/or buy a Mac mini. Problems solved: • The instance never sleeps. The agent just runs.
English
52
54
1.5K
282.8K
Adam
Adam@adamdotdev·
Programming was deeply satisfying work to me. Work for hours/days before getting the payoff of the code working well on your machine. I’m feeling so much friction now to open the editor and do this kind of task by hand, but also increasingly depressed with the nature of work in an AI assisted dev workflow. Back and forth prompting seems to eat at my soul. Need to find a balance that brings back some of the toil.
English
254
224
3.8K
502.6K
Beer van der Drift
Beer van der Drift@BeerDrift·
@levelsio Working harder helps, but is that the major problem here? As I understand it Germany is more in a productivity crisis (returns per hour worked) than a labor crisis (total hours worked).
English
0
0
0
196
Beer van der Drift
Beer van der Drift@BeerDrift·
Fascinating. Winners optimize for verification. I see this daily with coding tasks. The more tools I give my agent to verify it's work, the better the quality, while the underlying model stays the same.
Christian Catalini@ccatalini

1/ Some Simple Economics of AGI—🔥🧵 Right now, there is a low-grade panic running through the economy. Everyone is asking the same anxious question: what exactly is AI going to automate, and what will be left for us?

English
0
0
1
41
Beer van der Drift
Beer van der Drift@BeerDrift·
@levelsio What about both? I use CC daily for coding, but use Openclaw as my generic AI agent because of its plug and play in regards to cron, browser and channel use. They compliment each other, not replace (at defaults). Or do you see it different?
English
0
0
0
128
@levelsio
@levelsio@levelsio·
OpenClaw might be the gateway drug for normies to finally start using Claude Code like the rest of us
Daniel Sempere Pico@dansemperepico

I've been using OpenClaw exclusively as my only AI tool for the last three weeks. For content, I have an agent on Gemini 3.1 Pro for everything except social media posts where I have a dedicated agent on Opus 4.6. But today I hit a wall. For the last week I've been working on a project with my Gemini 3.1 Pro agent that required referencing several markdown files with ±100,000 words of source material that I either wrote or spoke and transcribed with speech-to-text. I've been sparring with the Gemini agent, going back and forth, refining sections, and frankly making painfully incremental progress. Then on a whim, I fed the source material into a project in the Claude Chat app on macOS and picked Opus 4.6 Extended to see if it could do a better job. My jaw hit the floor. It one-shot an absolute masterpiece. Opus 4.6 Extended took my voice and ideas from the source material, structured and polished it in a way that would have taken me 15 hours of back-and-forth with my Gemini agent on OpenClaw. The output was 97% of the way to a final draft, and it only took one messy spoken prompt plus the source material to get there. I don't know what wizardry Anthropic are doing, but it's working. (There's also a chance I was experiencing context rot with this task on Gemini 3.1 Pro through OpenClaw.) Regardless, I had to pick my jaw up off the floor. My use of Opus 4.6 through OpenClaw hasn't been as impressive as this experience with Opus 4.6 Extended in Claude Chat. I wonder if Anthropic is keeping some sort of special firepower for its own apps. If that's the case, I'm tempted to move to a full Claude Code setup, because the difference in this use case was that significant. Having the ability to think through and use such insane volumes of source material opens up a world where you can synthesise and condense unfathomable amounts of information into finished content in a few seconds. We're living through the golden era.

English
118
16
727
114.5K
Beer van der Drift
Beer van der Drift@BeerDrift·
@venturetwins My agent's memory literally is full of mentions of me getting frustrated with its actions 🙈.
English
0
0
0
9
Justine Moore
Justine Moore@venturetwins·
Noticed a strange phenomenon: the better a model gets, the more likely I am to get really frustrated. When a model passes my initial tests, I get excited and start thinking it can do everything. And then I have to crash back down to reality 🫠
English
33
1
133
8.1K
Beer van der Drift
Beer van der Drift@BeerDrift·
@jacob_posel Running from a local PC helps a lot. Don't use headless, otherwise add puppeteer-stealth plugin (works with playwright too).
English
0
0
0
48
Jacob Posel
Jacob Posel@jacob_posel·
What's the best and most reliable way to give your OpenClaw bot access to the browser?
English
26
0
26
11.5K
Beer van der Drift retweetledi
Richard Poelderl
Richard Poelderl@richardpoelderl·
Announcing Build & Rave Berlin on 13 March! Inspired by a Berlin co-working meetup where Pete demo'd his early version of @openclaw. Work is changing with AI and we want to bridge the gap between AI innovation & adoption together with @v0 + @ceciaramitaro & @elevenlabs 🧵↴
Richard Poelderl tweet media
English
1
3
15
1.5K
Beer van der Drift
Beer van der Drift@BeerDrift·
@loshmi In little ways! Scraping for upcoming events in my city. It aggregates and suggests without requiring setup. It helps me claim delayed flights by me just uploading a boarding pass saving time. But complex long-state work still leads to disappointing results.
English
0
0
0
334
Loshmi
Loshmi@loshmi·
to all people using ClawBot: did u create anything that actually helped you with your life?
English
275
5
392
49.7K