IVProduced

65.8K posts

IVProduced banner
IVProduced

IVProduced

@ivproduced

Just tweeting what I'm seeing. | AI Builder and Enthusiast.

Katılım Haziran 2009
2K Takip Edilen2.2K Takipçiler
IVProduced
IVProduced@ivproduced·
@_jaydeepkarale 90% of companies shouldn’t allow those. This is something you should do on your personal computer, but times are changing
English
0
0
0
49
Gnericvibes
Gnericvibes@GnericGladhizA·
@claudeai I hope Claude doesn’t end up sending someone’s nude as emails someday 😂
English
1
0
3
125
Claude
Claude@claudeai·
You can now enable Claude to use your computer to complete tasks. It opens your apps, navigates your browser, fills in spreadsheets—anything you'd do sitting at your desk. Research preview in Claude Cowork and Claude Code, macOS only.
English
4.6K
13.4K
129.8K
64.4M
IVProduced
IVProduced@ivproduced·
Funny thing is, I was making an openClaw before openClaw but mine wasn’t as good. It couldn’t code for me but it could: Manage my plex library Mange home automations Search the internet (not well) And send me news rss pushes I have been meaning to implement openClaw features but have been working on other things
English
0
0
0
75
Burke Holland
Burke Holland@burkeholland·
@ivproduced DO IT. You won't regret it. Its the one thing every develop should be doing rn - building their own claw to work their way.
English
1
0
0
84
NewsWire
NewsWire@NewsWire_US·
JURY SAYS MUSK DEFRAUDED TWITTER INVESTORS BEFORE 2022 BUYOUT
English
393
3.7K
25.6K
2.5M
IVProduced
IVProduced@ivproduced·
@theo We don’t let it go over an hour on its own…wtf lol
English
0
0
0
6
Theo - t3.gg
Theo - t3.gg@theo·
Just let Opus go for over an hour on a new feature. When it was done, I asked how I can test it. 20 minutes later, it realized I can't test it because it did the whole thing entirely wrong. Idk how you guys use this model every day for real work 🙃
Theo - t3.gg tweet media
English
431
32
1.6K
328.4K
IVProduced
IVProduced@ivproduced·
Was literally telling a senior engineer how you can’t to this in vs code, now I can tell him he can do it in 2 months
Pierce Boggan@pierceboggan

New in @code Insiders: Control reasoning effort from the model picker.

English
0
0
0
50
Pier Jarae 💅🏽
Pier Jarae 💅🏽@UrbanBlackGirl·
In another lifetime, somebody remind me to get put to sleep to get my wisdom teeth removed. 😭😭
English
1
0
0
70
Burke Holland
Burke Holland@burkeholland·
Been sparing with GPT-5.4 mini (high reasoning) in the @GitHub Copilot CLI for a day now and here's my take. The Good: MUCH better at following instructions and tool calling. Better even than Haiku. Fast. Rivals the speed of Opus 4.6 high reasoning which is crazy fast in Copilot. Shockingly - has pushed back on me two times when I asked it to do things that made no sense or where it didn't have enough context. Uses skills! Without being told! What a time to be alive. Does work well with plan / autopilot modes in the CLI. Calls ask_user tool when needed. The Bad: Still not so smart. Struggles debugging simple things like CSS. Having it explain the problem and then fix it seems to help a lot. Your brain is going to be required with this one. I don't feel like I'm getting improved accuracy in high and xtra high vs medium (default). It definitely thinks more, but it's not thinking very smart thoughts. I feel like the more it thinks the more it gets confused. Overall take: Solid budget driver. WAY better than anything we've got so far at this price point (.33x). Worth your time to invest in adding it to your workflows. And here is a 3D snek game it built, pushed to GitHub and deployed. Took me about 25 turns with the agent to get it to this point from the initial prompt. burkeholland.github.io/sneks/
English
5
2
56
6.3K
IVProduced
IVProduced@ivproduced·
@burkeholland I wonder if this is why people complain about anthropic models being “dumb” a few days after release. I have never had such issues
English
1
0
1
166
Burke Holland
Burke Holland@burkeholland·
I keep trying to tell people this. The context window is not a memory. It’s a room. The more stuff you put in there, the more cluttered it gets until eventually the model just stays confused. Don’t listen to me. Listen to Matt.
Matt Pocock@mattpocockuk

Doing some experiments today with Opus 4.6's 1M context window. Trying to push coding sessions deep into what I would consider the 'dumb zone' of SOTA models: >100K tokens. The drop-off in quality is really noticeable. Dumber decisions, worse code, worse instruction-following. Don't treat 1M context window any differently. It's still 100K of smart, and 900K of dumb.

English
18
13
146
16.8K
IVProduced
IVProduced@ivproduced·
Aye haiku smokin
Kyle Daigle@kdaigle

Hot take from looking at @github Copilot telemetry: benchmarks make coding models look wildly different. Production workflows make them look much more similar. 👀 We looked at 23M+ Copilot requests and examined one simple metric: code survivability.

Indonesia
0
0
0
125

Keşfet

@_jaydeepkarale @GnericGladhizA @claudeai @_CryptoKrew_ @DonaldJTrumpJr @TheInverseTimes @worldlibertyfi @NewsWire_US