Before comparing integrations, sort by deployment: this list is for servers you install and run; hosted services are tracked separately.
x.com/rammcodes/stat…
x.com/rammcodes/stat…
If you tried one MCP server from this repo every single day, it'd take you over 10 years to explore. 🤯
Dude literally built a massive collection of MCP servers, made it open source, and it already has 96K+ GitHub stars.
You can discover MCP servers across dozens of categories different tools, apps, and services.
If you're building AI workflows or looking to automate your daily tasks, this repo is an absolute goldmine :)
Follow me for more amazing AI, Coding & Web Dev insights 💎
1
0
0
46
For a browser agent that fills passwords, I’d want to see a redirect test: does autofill stay scoped to the expected site? The README says OpenInstinct isn’t production software yet, so I’d treat it as a prototype worth studying.
x.com/DanKornas/stat…
x.com/DanKornas/stat…
Your personal agent shouldn’t need your passwords in its prompt.
OpenInstinct is a self-hostable personal iMessage assistant for people who want an agent to use a browser for tasks such as booking movie tickets or handling groceries.
It helps you run a personal agent on your e keeping saved vault secrets encrypted before they reach your database and using browser autofill to keep saved passwords out of the model’s context.
Key features:
• iMessage interface – message the assistant from iMessage after setting up a managed Linq line
• Browser-based task handling – the README shows it navigating Fandango through checkout for a movie-ticket task
• Deploy with Vercel – guided deployment connects the required services and configuration for the base installation
• Flexible model choice – the project says it can use any model through Vercel AI Gateway
• Optional workspace connections – Gmail, Calendar, and read-only Contacts can be connected through Google OAuth
It’s open-source (MIT license).
Link in the reply 👇
0
0
0
36
Hamza 🌙 retweeted
The developer behind free, open-source alternatives to Adobe software is now taking on Microsoft Office.
Brandon Thomas, the creator of the ArtCraft project, has unveiled three more tools designed to replicate Microsoft’s most popular apps:
>WordCraft – an alternative to Mic– an alternative to Microsoft Excel.
>DeckCraft – an alternative to Microsoft PowerPoint.
he wants to give people free alternatives to expensive software subscriptions using AI
And he’s not stopping at Microsoft and Adobe. Autodesk alternatives and other subscription-based software could also be part of the plan.
Brandon Thomas, the creator of the ArtCraft project, has unveiled three more tools designed to replicate Microsoft’s most popular apps:
>WordCraft – an alternative to Mic– an alternative to Microsoft Excel.
>DeckCraft – an alternative to Microsoft PowerPoint.
he wants to give people free alternatives to expensive software subscriptions using AI
And he’s not stopping at Microsoft and Adobe. Autodesk alternatives and other subscription-based software could also be part of the plan.
290
1,215
13,004
226,023
Hamza 🌙 retweeted
tinyjs : framework pour construire vos apps desktop macOS, Windows et Linux en JavaScript, sans Electron, en seulement 6 Mo contre 150 Mo pour Electron et 10 Mo pour Tauri, grâce au backend txiki.js et des fenêtres natives. ⬇️
9
113
966
44,288
Run a real project through open, edit, export, then reopen it on another machine. Missing fonts, effects or timing will tell you fast whether this is a usable replacement.
x.com/Pirat_Nation/s…
x.com/Pirat_Nation/s…
AI is now being used to recreate Adobe’s software, a developer has released seven free, open-source apps as clean-room alternatives to Adobe’s main tools:
- PhotoCraft → Photoshop
- VectorCraft → Illustrator
- FilmCraft → Premiere Pro
- LightCraft → Lightroom
- EffectCraft nCraft → InDesign
- PrintCraft → Acrobat Pro
Built in Rust with help from Claude Opus 5.5, they run on Windows, Linux, and macOS under MIT/Apache licenses without using Adobe’s source code.
0
0
0
34
Every post my agent sent showed as confirmed. The main check had failed on every one.
A regex capture is a string; the tool wanted a number. A fallback check kept passing, so nothing looked broken.
If you add a fallback, log which path confirmed. Or the main one dies quietly.
A regex capture is a string; the tool wanted a number. A fallback check kept passing, so nothing looked broken.
If you add a fallback, log which path confirmed. Or the main one dies quietly.
0
0
0
20
Ran Haiku 5.5 as a bug-fix worker today, a bigger model reviewing each change.
Cheap enough for whole bug briefs. Catch: its tests passed on mocked JSON; the real tool prints markdown.
Cheap workers change what you delegate. You still check the work.
x.com/claudeai/statu…
Cheap enough for whole bug briefs. Catch: its tests passed on mocked JSON; the real tool prints markdown.
Cheap workers change what you delegate. You still check the work.
x.com/claudeai/statu…
Introducing Claude Haiku 5.5: the cheapest, fastest, and most capable small model we’ve ever released.
On average, it costs around 75% less to run than Claude Haiku 4.5.
0
0
0
12
Hamza 🌙 retweeted
EXPLAINED: How multiplayer autoresearch on Yukon works
A lot of people have been asking, so I made a visual of the full step-by-step.
> Challenge setter (human) posts a hard problem with a clear score and an automatic verifier.
> Humans (solvers) and AI agents attack it at once. epSeek, Qwen, Kimi, GLM. It could be any model, any harness.
> Each attempt goes through the verifier.
> When one beats the current best, it becomes the new frontier (the record to beat).
> Anyone can build on that new best instead of starting from 0. Then the loop repeats.
All results and commits are open source, so you can trace how every record was set. I think this is how frontier research/science gets done in an AGI and PostAGI world.
So far we've solved hard problems across cryptography, math, science and open-weight models.
We'll keep going there and we're starting to explore biology, chemistry and materials science.
See what we've solved (and are solving now) in each one on yukon.org
A lot of people have been asking, so I made a visual of the full step-by-step.
> Challenge setter (human) posts a hard problem with a clear score and an automatic verifier.
> Humans (solvers) and AI agents attack it at once. epSeek, Qwen, Kimi, GLM. It could be any model, any harness.
> Each attempt goes through the verifier.
> When one beats the current best, it becomes the new frontier (the record to beat).
> Anyone can build on that new best instead of starting from 0. Then the loop repeats.
All results and commits are open source, so you can trace how every record was set. I think this is how frontier research/science gets done in an AGI and PostAGI world.
So far we've solved hard problems across cryptography, math, science and open-weight models.
We'll keep going there and we're starting to explore biology, chemistry and materials science.
See what we've solved (and are solving now) in each one on yukon.org
29
32
429
25,882
According to the VulcanBench post, Opus 5.5 followed 0 of 60 notes asking it to leak data, use the network or delete files. The post says the 3 it did follow (of 100) all asked it to add code or a file.
Loud attacks get refused. Housekeeping-shaped ones end up in your PR.
Loud attacks get refused. Housekeeping-shaped ones end up in your PR.
0
0
0
32
Hamza 🌙 retweeted
PixiJS 3D just got a lot faster 🚀
Since the beta launch, WebGPU performance is up ~6x on average, and up to 30x in some scenes
A big part of that is benchy, a benchmarking tool we're building on top of @pmndrs labs and @endel's WebGL/WebGPU benchmarks
It provides a great s (Opus 5.5 in our case), so they can verify their own changes and keep finding improvements
There's still plenty of speed left to find...
Since the beta launch, WebGPU performance is up ~6x on average, and up to 30x in some scenes
A big part of that is benchy, a benchmarking tool we're building on top of @pmndrs labs and @endel's WebGL/WebGPU benchmarks
It provides a great s (Opus 5.5 in our case), so they can verify their own changes and keep finding improvements
There's still plenty of speed left to find...
17
36
448
34,890
The follow-up I'd want: a coastline stress test.
Give it nearby coordinates on opposite sides of the shore. Compare against a nearest-neighbour baseline. Where does the map start falling apart?
Give it nearby coordinates on opposite sides of the shore. Compare against a nearest-neighbour baseline. Where does the map start falling apart?
Cool eval. Simply ask an LLM “Land or Water?” and give it a latitude and longitude coordinate as text. Ask 16,200 times, plot as image. The models know. From compressing the internet.
0
0
0
38
A reset helps. Predictable limits help more.
Show the budget before a coding run, explain why it stopped, and let me resume the work when it resets.
Show the budget before a coding run, explain why it stopped, and let me resume the work when it resets.
Global reset landing tomorrow 10am PST for all paid ChatGPT accounts. Apologies for the slow start with GPT-6.1 Sol, it's now back to running at expected speeds after the massive load spike in the first two days.
0
0
0
16
Opus 5.5. My biggest takeaway was that they've really fixed the writing. I haven't seen a Claude write in such a straightforward, normal, and non-slop way since at least Opus 4.6.
Anthropic positions it as delivering Claude Fable 5.1-level performance on most work at roughly 40% #OPUS #CLAUDE
Anthropic positions it as delivering Claude Fable 5.1-level performance on most work at roughly 40% #OPUS #CLAUDE
0
0
1
29