shaped

2.2K posts

shaped banner
shaped

shaped

@shaped

Post training data for RL | prev 2x exits

San Francisco Katılım Mayıs 2009
210 Takip Edilen305 Takipçiler
shaped
shaped@shaped·
@andrewjclare Well Apple themselves draw inspiration from Braun and Dieter Rams' philosophy, I guess it's all about stealing like an artist
English
0
0
0
405
Andrew Clare
Andrew Clare@andrewjclare·
crazy how much apple influences other brands. no one did the camera plateau like this until apple. they continue to be “the brand”
Andrew Clare tweet media
English
240
40
1.7K
243.8K
shaped
shaped@shaped·
@localghost For the people saying this is just an ipad with a keyboard, yeah but I think as a programmer I'd rather buy this than the ipad thanks to macOS and the zsh terminal
English
0
0
1
38
shaped
shaped@shaped·
I dunno man, I think I would rather deal with macOS than with ipadOS when it comes to how walled the garden is. I think this would be a pretty sweet deal for indie devs who want to test out their swift apps using xcode for cheap and still be able to install open source packages, use the terminal and have root access and stuff like that
English
0
0
0
75
shaped
shaped@shaped·
Steve Jobs used to just sit on a pile of cash during his time at Apple for "rainy days" and I suppose its part of the company culture to be highly liquid at all times to weather storms as they come. Freezing their RAM prices implies that they can and will ride out this hike, which is reassuring that a big tech company believes RAM prices will return to normal
English
0
0
0
83
Alex Cheema
Alex Cheema@alexocheema·
Nobody is talking about @apple keeping prices the same for the 128GB MacBook Pro. There has been no price increase in response to surging memory prices. Everyone is talking about the boost in compute, speeding up prefill by 4x. This is cool but practically it’s not that big of a deal. Why? Because on your own computer, most apps/tools using LLMs are going to get high kv cache hit rates - that means as a user you only experience slow prefill once. kv cache can be persisted to disk and loaded at 6GB/s. Most time in LLM inference is spent on decode, which is memory bandwidth bound. It’s still great for image/video generation, high batch LLM inference and fine-tuning, which are compute bound. We should see huge speedups there. Apple’s AI strategy is on-device LLMs and here, memory is the name of the game, not FLOPS. Expect the same for M5 Pro/Max Mac Mini and M5 Ultra Mac Studio. That means 512GB M5 Ultra at 10k! @tim_cook is a supply chain genius.
Greg Joswiak@gregjoz

The all-new MacBook Pro with M5 Pro and M5 Max pushes the boundaries of what you can accomplish from anywhere. Run advanced large language models on device and unlock capabilities that can't be done on any other laptop—all while maintaining exceptional battery life!

English
45
37
584
142.4K
shaped
shaped@shaped·
@linguinelabs Yes, this is what I do all the time! Starting to run out of memorable ports and ram at this point
English
0
0
1
1.5K
Kevin
Kevin@linguinelabs·
Pretty insane that I can just vibe code any web app I want (acne tracker, gym tracker etc), run it on a port of my laptop, and then access it on my phone via TailScale. No need to deploy to a domain or app store, all free, and completely tailored to me.
Kevin tweet mediaKevin tweet media
English
56
135
3.4K
227.8K
shaped
shaped@shaped·
@zekramu What goes in is what comes out. We put in truckloads of code and we expect an engineer on the other side, but all we get is a code machine instead
English
0
0
1
25
zek
zek@zekramu·
basically all coding models are good enough to do whatever is asked of them, and simultaneously are terrible at handling ambiguity. when left to make choices all of them make the easiest & often worst choices. such that they might be good programmers but terrible engineers.
English
25
6
235
13.2K
shaped
shaped@shaped·
This just puts it up for interpretation and may waste court time in the future. What's to say (or argue against) that it is meaningful modification to AI generate a mascot and edit said mascot into my logo and branding elements? I did AI generate the mascot, but I did meaningfully modify my logo with conventional photo editing tools to incorporate mascot into it
English
0
0
0
76
Tim van Helsdingen
Tim van Helsdingen@TvanHelsdingen·
So many people are misinterpreting this. You can still copyright AI content, but only if you meaningfully modify it, which is what every studio will do anyways. Still a win on the breaking down copyright law though (since copyright law definitely needs a revision)
ToonHive@ToonHive

Ai-generated artwork officially is ineligible for copyright protection as the US Supreme Court declined to review a appeal case. The court rules that artwork needs to have a human creator in order to be eligible. (Source: reuters.com/legal/governme…)

English
24
5
63
5.5K
shaped
shaped@shaped·
The main reason why AI writing is even such a hot topic is because AI just lets anyone use it as they please and the result is a physically similar article. A very inexperienced writer and an experienced writer both would produce an article of 500 words given a topic. But the inexperienced guy would simply just ask the AI to "write something on X" thus producing the purest slop known to mankind, vs the experienced person taking their time to collect their thoughts and concepts and coherently putting them together and filling in the blanks with AI, producing something worth reading as opposed to slop But people see "made with AI" and their first thought is the inexperienced guy and they immediately say "this is AI slop, I'm not reading this" instead of examining if the concepts exposed is slop or not
English
0
0
2
36
shaped
shaped@shaped·
I mean, the macbook mini would sound like it was physically smaller, which it isn't, and they can't call it just a macbook because the macbook air implies its thinner than the base macbook (it's not) and its already established that a macbook is always one of air or pro and its cognitively easier to grasp neo that way
English
0
0
1
49
shaped
shaped@shaped·
@Barney_H_Y Too many vibeshifts in the last 4 months, I'm just very glad the APIs are interchangeable or else it would be very ugly shifting from one model to another as much as I do right now
English
0
0
0
282
shaped
shaped@shaped·
@thekitze They just shipped an ipad with an attached keyboard and called it a day 😭 designed for educators and it doesn't even have an HDMI port to let professors connect to projectors without a dongle, questionable position for the audio jack so many things wrong with this
English
1
0
1
178
shaped
shaped@shaped·
@jparkjmc I mean, a mac mini is the same price so i guess ultimately it comes down to if you'd rather have a screen or more ram for your openclaw. I'd still pick ram because I usually have my claw run multiple subagents parallely and that uses up a lot of ram.
English
0
0
0
219
shaped
shaped@shaped·
Wow this is the absolute worst case. The ideal best case would be for both agents to realize that they're talking to an agent (preferably using an imperceptible smell, style of talking or an audio that plays early in the call or other methods which lets the agents know theyre agents) and them shifting to a very low token communication method to achieve the intended outcome
English
0
0
3
138
Olivia Moore
Olivia Moore@omooretweets·
New voice agent edge case just dropped 😂 “They kept politely confirming things, asking for clarification, thanking each other, re-confirming previous confirmations”
Olivia Moore tweet media
English
22
4
103
9.3K
shaped
shaped@shaped·
@aaronp613 Was really holding my breath on this one
English
0
0
0
268
Aaron
Aaron@aaronp613·
IMPRESSIVE: Apple did the impossible Apple's Polishing Cloth is confirmed to work with the MacBook Neo
Aaron tweet media
English
130
338
9.3K
340.6K
shaped
shaped@shaped·
@chrysb @openclaw openclaw's webUI is one of the most underutilized things of the repo entirely. You could do a lot of setup through the webUI that you didn't have to painstakingly chat with the agent/go through the CLI to set it up for you. I wish it were better designed like this
English
0
0
0
91
Chrys Bader
Chrys Bader@chrysb·
sneak peek @openclaw + google workspace, minus hours of config
Chrys Bader tweet media
English
28
7
264
21.7K
shaped
shaped@shaped·
@darkmembo API calls have been a very longtime common metric tho
English
0
0
0
343
Mark Dembo
Mark Dembo@darkmembo·
First time I’ve seen MCP tool calls as pricing metric.
Mark Dembo tweet media
English
11
0
105
17.4K
shaped
shaped@shaped·
@dreamingtulpa I can't call these things slop anymore 😭 I'm really enjoying watching these
English
0
0
1
33
shaped
shaped@shaped·
@altryne This was actually a nobrainer feature coming in. The friction to disable all the permissions vs the friction to disable only the unsafe permissions was crazy, I'm glad it has gotten so much better now. Cant imagine having to make all those complex hooks anymore
English
0
0
8
655
Alex Volkov
Alex Volkov@altryne·
Claude Code is about to launch a "--less-dangerously-skip-permissions" mode AKA "auto mode" 👀 Given the huge enterprise adoption this makes a lot of sense!
Alex Volkov tweet media
Daniel Sternlicht@dsternlicht

Was about to run #ClaudeCode with "--dangerously-skip-permissions" when @AnthropicAI emailed me a safer alternative at that exact moment. Like grabbing junk food at midnight and getting your weekly health report notification. Still hit Enter though. auto mode can have me tomorrow 😁

English
17
19
609
80.7K
shaped
shaped@shaped·
@bbssppllvv So this is just mobbin as an MCP 😭 stop wiping our "taste" UI moats man
English
0
0
0
382
Mike Bespalov
Mike Bespalov@bbssppllvv·
Imagine your AI doing design research before generating UI. Studying real screens and user flows instead of guessing That’s Refero MCP refero.design/mcp
English
43
111
1.9K
147.5K
shaped
shaped@shaped·
I like Glaze. You don't always like dealing with a chat window and wasting tokens with an agent to get some basic things done like with openclaw (altho it might be cognitively less load sometimes) so its better to chat with AI once to get an easy app done and installed into your computer as a one time process instead
English
0
0
0
196
Dan Hollick
Dan Hollick@DanHollick·
Been using Glaze for a few weeks to make personal tools and its so incredibly slick. Made this image converter today that watches folders and converts / resizes any images added to it based on some rules. Something only I need, tailored to my exact workflow.
Raycast@raycast

Today we're launching Glaze 💠 Create any desktop app in minutes by chatting with AI. Beautiful, powerful, and truly personal. Learn more on glazeapp.com Follow @glazeapp for updates.

English
27
22
744
92K