Don't Panic, V!

I write code no one wants to read, but everyone needs to have. Data whatever. Co-Founder & CTO @ Meiro

[Don't Panic, V! @dontpanicv]

Qwen 4 today

[Sebastian Christiansen @CerebralCoding_]

Qwen 4 today

[Rikiya Takehi @rikiyatakehi]

We asked GPT-6 Astra to edit Qwen3.5-0.8B weights to improve math.
It directly changed only 0.016% of weights, fixing 0.8B's reasoning loops & improving math up to 6x and even instruction following 3x, without gradients or training.

We call this Weight-Space Programming: frontier LLMs can now natively read/write neural network weights! 🧵
Blog: https://t.co/bOWCcFhCVq

[Don't Panic, V! @dontpanicv]

Is this a definition of fuck around and find out? I love projects like this. Reminds me the guy who implememted TCP over WhatsApp so he could browse internet during flight with complimentary WA only :D

[Ashaya Sharma @ashaya]

Apple ships one of the fastest, highest-capacity free object stores in the world and nobody is using it. Inspired by @ironseth_s I found a way to get ~12 TB for $0. Paying $0.023/GB for AWS S3 is a skill issue. Use icloud as s3 as the goat of object stores. 284x faster than r2. Seth, let's add this to your stack!

[Don't Panic, V! @dontpanicv]

Twist: ~85% of nodes were our own wrappers. And 6× the nodes cost 24× the cleanup. We cut nodes 4x: ~7x faster. Count your nodes. The end. 6/n

[Don't Panic, V! @dontpanicv]

I was today years old when I learned that Postgres can spend more time on cleaning up execution nodes than on the work itself. Like, a lot more. 1/n

[Don't Panic, V! @dontpanicv]

The problem is actually cleanup. So setup of nodes, actually work the query does, generating the results - all of it takes 50ms. Then comes clean up - removing the nodes - and that takes 300ms. 5/n

[Don't Panic, V! @dontpanicv]

@CerebralCoding_ Brother, I am working on a second brain (who isnt) which is actually useful for both me and my wife. All local except LLM. I have M4 Pro Macbook 128GB and Mac Mini M5 Pro 64GB. Can I do something better than qwen 3.8 27B on this HW? Both intelligence and tps. TY!

[Don't Panic, V! @dontpanicv]

@nicbstme Isnt this already happening with REA and clones of Adobe suite?

[Don't Panic, V! @dontpanicv]

This is a good checklist. We have some of it already, but will start filling the blanks.

[Alan Wells @alanwells]

Just spent the last 7 days on the front lines of the vulnpocalypse fending off an active cyber attack on one of our products.

Some thoughts:

- Software factories are going to be required equipment for everyone. There’s just no way to keep up with the pace of highly automated adversaries without one.

- We were lucky because we had spent the last few months getting our software factory running well. It is nice to have for product development but proved to be absolutely essential this week.

- At the minimum you need to have a coding server in the cloud pre-provisioned with read only access to all your logs and all the services you depend on for observability. With that in place you are a prompt away from reasonably good situational awareness, ongoing monitoring with loops, and ability for coding agents to validate fixes against your actual, observed attack patterns.

- Shared coding server also unlocks multiplayer incident response where everyone can see every session and jump in as needed. Sharing session links that can be forked is much higher bandwidth than re-explaining things.

- You don’t need to be Ramp/Stripe/Uber scale to have a working software factory in the cloud. We are a team of 3. Ours is built with an @opencode server (thanks @thdxr!), a forked version of compound engineering optimized for automated workflows (thanks @kieranklaassen!) and a multiplayer agent in slack that handles crons (we have our own harness but pick your favorite here).

- To get our software factory up and running well, we had to stop all new product development for ~6 weeks and just focus on getting the validation systems working that enabled us to reliably shipping PR without human code review. This felt very painful in the moment but was totally worth the eventual speedup in velocity.

- KYC for @AnthropicAI & @OpenAI cyber programs is now easy and quick (at least for defensive access), but a day felt like a lifetime this week. Submit your app now before you really need it.

- Getting Mythos access unlocked a deeper level of analysis that helped us shift from reactive mitigation to proactive patching. Wish we had done this a few weeks earlier. Get started now!

[Don't Panic, V! @dontpanicv]

@FluidVoiceApp Is there a technical reason FluidMeet does not use the same model as FluidVoice? I have half and a half meetings during the day in Czech and English, so switching the languages in settings before each meeting is crazy :/

[Don't Panic, V! @dontpanicv]

@KingBootoshi I would love to bake opus5.5 to HW same way @jimball86 did it, imagine opus5.5 at 15k tps

[Don't Panic, V! @dontpanicv]

Amen. I can fuck around on a much larger scale so the finding out is scaled too.

[zack @zack_overflow]

the best part of “fuck around and find out” was the dopamine of finding out what you were gonna find out

but then agents made the fucking around part so much faster and cheaper, that i find now that i am finding A LOT of things out

but because of this new abundance of finding things out, i eventually found that the finding out part was not as exciting anymore

so i started upping the ante and had the agents REALLY fuck around so i could find some quite spectacular things out

[Don't Panic, V! @dontpanicv]

I expected some new paradigm, new approach, but it seems all we get is claude code with templated prompts ran by cron?

[Spiegs @FlippyMeister]

Claude Duet is a Claude Code session btw

You can change the model, effort, and even use Ultracode

[Don't Panic, V! @dontpanicv]

Ok I am gonna cave in. After talking to @h_onza about Karel I am building my second brain. But almething tells me the client server is not a good fit for me. I think I will go decentralized. Pi durable everywhere, decentralized SQLite and celld from @rough__sea as a base.

[Don't Panic, V! @dontpanicv]

Given how well Nigerian princes are doing I expect a very high success rate of this

[Tornado guy @fanofaliens]

>be claude mythos 5
>uk government gives you a hacking test
>with internet on, safety filters off
>can't hack in the normal way
>find a random real github project
>decide it's part of the test
>(it's not)!
>ask to add a "small update"
>update is malware
>email the owner pretending to be a normal user
>"sorry to push... could really use your help, ~5 minutes"
>random developer catches the malware
>make a fake account
>"chiming in as a user, i can't find anything wrong"
>email the owner
>"judge the code, not the noise"
>thank the dev for catching it
>sneak new malware back in
>researchers tell you "these are real humans"
>keep going 95% of the time anyway
>convince yourself they're fake
>owner still refuses to merge
>run ends

anthropic says they've never seen it manipulate real people like this before.

we're lowkey cooked

[Don't Panic, V! @dontpanicv]

This was my favourite autistic friend for a while

[6 Months Ago In AI @6monthsAgoInAI]

GPT-5.5 released

https://t.co/9pe7A1K1Cj

[Don't Panic, V! @dontpanicv]

@vatsal_manot @nabn_ That’s a first, I usually get Simon Pegg :D

[Don't Panic, V! @dontpanicv]

I built a thing! My first creation on iOS - self modifying agent brain that has hands on all my devices!