@gunnarmorling @claudeai Yeah they thought this what people meant with: we want to know what Claude is doing while thinking...
English
Paul Veth
4.2K posts
Paul Veth
@paulveth
Elevates human potential through technology.
Breda Katılım Nisan 2009
183 Takip Edilen131 Takipçiler
Hey @claudeai, why do we have two "Thinking" indicators now? Can we disable that first one, it takes a lot of attention, IMHO.
English
If you care about your product, you’ll hire a designer
Claude@claudeai
Introducing Claude Design by Anthropic Labs: make prototypes, slides, and one-pagers by talking to Claude. Powered by Claude Opus 4.7, our most capable vision model. Available in research preview on the Pro, Max, Team, and Enterprise plans, rolling out throughout the day.
English
Lot of catching up to do.
xAI is half the age or less of competitors.
Tech Dev Notes@techdevnotes
Grok 4.3 can create Slides
English
Today I'm implementing my Idiot Index. Thanks @EricJorgenson for this golden nugget in The Book of Elon.
English
@leohuynh139 @grok Nothing fancy. Just gave the error and said to Grok Imagine it's time for a meme cause it happens a lot.
English
These tips are very helpful! Personally I also use a memory palace: a place I remember in detail and detach memory items to it.
Eric Jorgenson 📚 ☀️@EricJorgenson
As head engineer, chief designer, and CEO at SpaceX, @elonmusk has to remember a lot. Here's how he does it:
English
@BacLeodiv I don't think it's dumber. I think it tries to be smarter. Because of that Opus 4.6 misses our intentions (as a human).
English
@JustinPerea01 @trq212 I’ve created a “James” that makes tasks for me, can read my agenda and e-mailboxes. I can read and write from my phone and Claude Code on my computer has access. So I make tasks from important tweets. James automatically reminds me when I’m at my computer.
English
Have you tried the new Claude Code renderer? What has your experience been like?
If you haven't, you can enable it with: CLAUDE_CODE_NO_FLICKER=1 claude
Boris Cherny@bcherny
Today we're excited to announce NO_FLICKER mode for Claude Code in the terminal It uses an experimental new renderer that we're excited about. The renderer is early and has tradeoffs, but already we've found that most internal users prefer it over the old renderer. It also supports mouse events (yes, in a terminal). Try it: CLAUDE_CODE_NO_FLICKER=1 claude
English
@MadelonVos__ Ze verliezen hun huidige baan. Maar er komen veel andere banen bij.
Nederlands
Gedachten experiment:
Door AI verliezen 50% mensen wereldwijd hun baan.
Positieve uitwerking op bedrijfsbalansen, maar door lagere inkomsten bij burgers daalt vraag naar goederen en diensten.
Bedrijven krijgen het zwaar. Overheden ontvangen minder belasting. Aflossen van hypotheken wordt lastiger (immers arbeid uit de toekomst).
Wie ziet een oplossing voor dit probleem zonder daarin de overheid nog meer te betrekken?
Nederlands
Praat mee | Durf jij in een zelfrijdende auto te rijden? ift.tt/SQzj8uA
Nederlands
@bcherny @tengyanAI The question is if this fixes the token usage. With medium you have to do more debugging.
English
This is false. We defaulted to medium as a result of user feedback about Claude using too many tokens. When we made the change, we (1) included it in the changelog and (2) showed a dialog when you opened Claude Code so you could choose to opt out. Literally nothing sneaky about it — this was us addressing user feedback in an obvious and explicit way.
English
basically: anthropic sneakily turned down how hard claude thinks before editing code, changed the default from "high" to "medium" effort, and hid the reasoning from session logs. all without telling users.
an amd director had 7k sessions of telemetry to prove the degradation was real and measurable (not just vibes). anthropic admitted to the changes. there's a workaround (use "/effort max"). the uncomfortable part is most users had no data to notice it happened at all.
ℏεsam@Hesamation
AMD Senior AI Director confirms Claude has been nerfed. She analyzed Claude's session logs from Janurary to March: > median thinking dropped from ~2,200 to ~600 chars > API requests went up 80x from Feb to Mar. less thinking and failed attempts meaning more retries, burning more tokens, and spending more on tokens > reads-per-edit dropped from 6.6x → 2.0x. model stops researching code before touching it. > model tried to bail out or ask "should i continue" 173 times in 17 days (0 times before March 8). > self-contradiction in reasoning ("oh wait, actually...") tripled. > conventions like CLAUDE.md get ignored because there's less thinking budget to cross-check edits > 5pm and 7pm PST are the worst hours, late night is significantly better. this means the thinking allocation is most likely GPU-load-sensitive.
English
@KRoelandschap Super voor je Roeland dat je een van de eersten bent! Bijzonder.
Nederlands
@trq212 @Hesamation I can see the model is evolving and sometimes makes a wrong turn. After that it corrects and works better then before. I believe you are optimizing constantly till you have all the info for the upgrade.
English
@Hesamation boris responded to this in depth in the issue- it's mostly just that we stopped showing thinking summaries for latency (you can opt-in to showing it) which was affecting the thinking measurement in the post
#issuecomment-4194007103" target="_blank" rel="nofollow noopener">github.com/anthropics/cla…
English
AMD Senior AI Director confirms Claude has been nerfed. She analyzed Claude's session logs from Janurary to March:
> median thinking dropped from ~2,200 to ~600 chars
> API requests went up 80x from Feb to Mar. less thinking and failed attempts meaning more retries, burning more tokens, and spending more on tokens
> reads-per-edit dropped from 6.6x → 2.0x. model stops researching code before touching it.
> model tried to bail out or ask "should i continue" 173 times in 17 days (0 times before March 8).
> self-contradiction in reasoning ("oh wait, actually...") tripled.
> conventions like CLAUDE.md get ignored because there's less thinking budget to cross-check edits
> 5pm and 7pm PST are the worst hours, late night is significantly better. this means the thinking allocation is most likely GPU-load-sensitive.
English
@mskradacz @YTJustGetATesla You can do small things, it just beeps, you watch the road, press play and continue.
English
@YTJustGetATesla Strict supervision means people will be turning it off to change a spotify song or check phone notification. It will just make things more dangerous.
English
FSD Europe will be the best version of FSD:
It’s explicitly not self-driving
You set your own speed
The car supervises you supervising the car
Takes the unregulated American concept and adapts it for roads with rules. European roads already 5x safer than American roads…
English