Matthew Spence (@mrspencehimself) header
Matthew Spence (@mrspencehimself) profile picture

🚀 Shipping Software and AI - see what I do 👇

United Kingdom
Joined February 2018
Photo by Matthew Spence (@mrspencehimself)Photo by Matthew Spence (@mrspencehimself)Photo by Matthew Spence (@mrspencehimself)Photo by Matthew Spence (@mrspencehimself)Photo by Matthew Spence (@mrspencehimself)Photo by Matthew Spence (@mrspencehimself)Photo by Matthew Spence (@mrspencehimself)Photo by Matthew Spence (@mrspencehimself)Photo by Matthew Spence (@mrspencehimself)
utterly huge progress - if you want a private mobile phone AI assistant, using local AI, works offline, code published on GitHub, you have got to check out Whisper AI (yes im associated) Find link on my bio website
Biggest update yet for Whisper AI: 🎙 Audio — record, transcribe and chat with your conversations. On-device. 🖼 Images — ask questions about what you see. 🤗Hugging Face — search and download models directly in app. 🔀 Multi-provider — Apple Intelligence, ChatGPT, Open Router and more. Switch between models mid-conversation. Some features require newer devices. Older hardware still gets a fully private AI assistant.
1
116
It's not often Fable fumbles
3
81
A self burn - now those are rare
Unpopular opinion: "AI makes everyone a developer" is true the same way "cameras makes everyone a photographer"
61
Matthew Spence retweeted
Cloudflare Mesh is here. Ready to connect your devices, servers, and agents to a single private network 🔐 And with Workers VPC, your Workers, Agents, and Durable Objects running on Cloudflare can now reach your private MCPs, APIs, and databases directly Oh, and it's 50 nodes + 50 users free on every account
47
190
51
1,705
281,977
Theory - the more relevant "measuring sticks" you provide an AI towards a goal, the less oversight you need
1
28
give me storage, of any size, and I will always fill it
1
36
My feed is so obviously 50% ai written now
1
1
115
Matthew Spence retweeted
Today @GoogleMaps is getting its biggest upgrade in over a decade. By combining our Gemini models with a deep understanding of the world, Maps now unlocks entirely new possibilities for how you navigate and explore. Here’s what you need to know 🧵
1,090
3,920
1,652
43,620
28,870,519
for Typescript developers out there when you store JSON locally, the structure inevitably changes over time we've been dealing with this ourselves across a few projects and figured it was time to build something small and focused to handle it properly. so we created a type-safe JSON migration library for TypeScript github.com/Nano-Collective/o…
1
3
94
Watching OpenAI sell out the public in real time
Today was a CRAZY day in the AI space. Morning - Anthropic CEO Dario Amodei refused to work with the Pentagon because they wanted to use Claude for mass surveillance and autonomous killer robots. Afternoon - OpenAI’s Sam Altman came out in support saying, “For all the differences I have with Anthropic, I mostly trust them as a company and I think they really do care about safety.” Evening - President Trump banned Anthropic from every federal agency in the United States government. Night - Sam Altman flipped. OpenAI submitted a bid to replace Anthropic and officially reached a deal with the Pentagon.
55
Matthew Spence retweeted
You can now write chat bot logic once, and deploy across Slack, GitHub, Teams, and Discord. We're open-sourcing a new Chat SDK, a unified TypeScript library to build for many chat platforms with a single codebase. Now available in public beta. vercel.com/changelog/chat-sd…
55
164
66
1,789
528,285
Visual benchmarks are just around the corner for Whisper AI App - then we can really demonstrate progress with our on-device private AI 🔥🔥
Fine-tuning tiny models is one of our strategies to bring private, local AI up alongside the models that come from big tech. The output of this research will bring direct value to both @UseWhisperAI and the software being built in @nano_collective. There's so much R&D to do in this space - every day you're learning.
2
84
Matthew Spence retweeted
Fine-tuning tiny models is one of our strategies to bring private, local AI up alongside the models that come from big tech. The output of this research will bring direct value to both @UseWhisperAI and the software being built in @nano_collective. There's so much R&D to do in this space - every day you're learning.
We're adding LLM-as-a-judge to Nanotune! The video below is Devstral 2 benchmarking the tiny 1.5B bash model we trained. More on all of this soon - the output of this research will hopefully mean we're offsetting more tasks to local, private models over ones from big tech 🔥
1
1
1
3
203
Building on-device Ai while sat in a British pub (yes the app is live!) Adding new stop functionality right now
2
7
134
Matthew Spence retweeted
Replying to @personofswag
it flipped from "write code" to "architect systems that AI writes for you while making sure it doesn't hallucinate your prod database into oblivion" harder and more fun honestly
4
11
1
344
22,104
Almost forgot to mention - we went fair-source with Whisper AI last week Now anyone can see for themselves that the AI on Whisper is 100% private on their mobile phone I'll put a link to the code in the comments here - DMs welcome
2
1
6
305