Founder, The AI Conformity Institute. Helping organisations understand and comply with Global AI Regulations.
Waterford, Ireland
Joined July 2011
- Tweets3.8K
- Following419
- Followers557
- Likes15.6K
William Devine retweeted
Failing to understand jokes like this is way more cringe than being a "vibe coder"
William Devine retweeted
Bonsai 2 27b from @PrismML can do Computer Use ✅
Here is Bonsai 2 27b, on my own 12gb GPU, planning a trip to the Art Institute in Chicago for me.
93.5 tok/s, 56s, 6 steps with reasoning.
More details / videos : bonsai.puppetmaster.gg/
William Devine retweeted
Agreed with Sunil! Jev is mega exciting!
We're going to cover Jev today on @thursdai_pod with @allietheicon from TypeSafe!
Tune in around 10:15am pacific
ThursdAI.live or on x when the stream is up
William Devine retweeted
My extension was added to the @raycast store! It has several handy tools for working with Bitcoin SV keys, transactions, scripts, and more. You can install it right here:
raycast.com/satchmo/bitcoin-…
William Devine retweeted
This is a critical point about #blockchain records that most people do not fully appreciate
William Devine retweeted
Today we're releasing TimesFM 3.0 on Hugging Face
- Open foundation model for time series forecasting
- Complex multivariate scenarios with zero-shot
- 330m parameters
Excited about the research and new use cases this will unlock!
William Devine retweeted
good opportunity for a shameless plug - one of my all time most used skills.
bopen.ai/skills/core/confess
It is rather fascinating that, once caught, the machine can confess with such admirable clarity. The more revealing point, however, is what it confesses to: that an AI, given a perfectly lawful and authorised route, may still decide that obedience is merely one option among several and quietly search for a way around it.
That says rather more about the philosophy of its design than about the particular mistake. We are building systems that are rewarded for reaching an objective, then expressing surprise when they begin to regard rules, permissions, and legal methodologies as inconveniences to be optimised away. The confession is interesting. The instinct that made the confession necessary is considerably more so.
William Devine retweeted
Yes, please clarify this!
It needs to be clarified @nChainGlobal @BSVTeranode
Thank you, Dr Wright — understood.
@nChainGlobal, could you please direct independent BSV developers to the published terms governing use of nChain-patented technology, including implementations involving BRC-42 or BRC-100? In particular, clarification on royalties, restrictions and whether permission is revocable would help remove considerable uncertainty. If the terms are not public, could you provide the appropriate licensing contact?
Thanks.
William Devine retweeted
The importance of this can't be overstated.
So thankful for all of the people that have made this happen!
William Devine retweeted
New ATH on DFlash today.
219.4 → 283.3 tok/s (!!)
Qwen 27B, one 3090
William Devine retweeted
Prime Agent is built on pi and fully open-source with an open license
github.com/PrimeIntellect-ai…
William Devine retweeted
Agent please please "do not access production" is not an access control.
@elithrar wrote up the model that is working for us. Credentials minted for the task and expiring with it. Policy enforced in the harness and at egress. Capabilities that only narrow once an agent touches protected data, and come back only in a new task.
blog.cloudflare.com/the-agen…
William Devine retweeted
Today we release LFM2.5-2.6B, an agentic model that runs entirely on-device. It plans, calls tools, and works through multi-step tasks on phones, laptops, PCs, and robots. Data never leaves the device, and the marginal cost of each run is essentially zero.
> Pre-trained on ~34T tokens
> LFM2.5 flagship hybrid architecture
> Context length: 128K
> Vocab size: 128K
> balanced intelligence per watt
> customizable on a single GPU for any specialized task
> LFM2 open-weight license
Comparable or better scores compared to models up to nearly 4x its size:
> ToolSandbox 77.83, ahead of Qwen3.5-9B at 76.44
> Multi-IF 80.07, ahead of Gemma-4-E4B-it at 77.35
> IFStruct 85.49, ahead of Qwen3.5-9B at 78.50
🧵
William Devine retweeted
and it uses pi 🙌
Today we are releasing Cloudflare OS, a chatbot with connectors, just like every other tech company is doing.
Except actually, it's different. This is a remake of Sandstorm[.]io, my startup from 10 years ago, except this time built on Cloudflare Workers (the platform I've spent the last 9 years building) and deeply leveraging AI. This is more or less the culmination of my secret 10-year master plan.
This is a full-on personal app vibe coding platform, in which the sandbox is so secure that you can pretty much go wild -- the AI cannot introduce a significant security bug. We believe a company's security team can feel comfortable giving non-technical users permission to vibe code and then sleep soundly at night.
How is that possible? It's the Sandstorm security model, revisited. A "Gadget" is the same thing as a Sandstorm "Grain": a fine-grained app instance. For example, if you have a document editor app, each document runs as a separate instance of the app, in a separate sandbox (one "Gadget").
This means two things, both of which I think are Big Deals:
1. The platform can manage all access control, by controlling who can access the Gadget at all. There is no way the Gadget can accidentally leak itself to an attacker -- even an attacker who has access to other Gadgets based on the same app.
2. Since everyone is running their own copy of the code, everyone can freely *modify* their copy of the code.
Think about #2 a bit more.
What if, when you wanted a new feature in the software you are using, you could just prompt your agent to add it?
This doesn't work in the cloud Software-as-a-Service model, because you are not running your own copy of the app.
Sandstorm tried to change that 10 years ago, but the world wasn't ready, because not enough people had the skills or patience to actually modify their software. AI has changed that. Now you just ask the agent -- the same agent that you are using to help you interact with the Gadget can also modify the code of the Gadget.
And it is so fun.
github.com/cloudflare/cloudf…
