Founder & CEO, Signal Eight. Passionate about AI & Tech. Ex Bloomberg
APAC
Joined June 2009
- Tweets1.3K
- Following590
- Followers186
- Likes2.4K
I won first place in the global OpenAI × @AITinkerers hackathon!!!
I built Sidekick to get on with the work while your meeting is still happening.
The idea came during the hackathon welcome video. I wanted an agent that could participate in a meeting and get on with the work as we talked.
Sidekick uses gpt live for voice along with the new agents api, @Exa for research and factchecking, and @ambiguousio so the team and agent can all work on the same documents.
Really proud of this one as the AI Tinkerers community is immensely technical and there were 700 teams participating! Still buzzing!
You talk it through. Sidekick gets to work.
I built an AI meeting teammate that researches and creates docs, slides & sheets while you talk.
I built this in just a few hours for the @AITinkerers global hackathon.
GPT-Live handles the conversation, while the OpenAI Agents API coordinates research and drafting. Ambiguous AI @ambiguousio is the shared workspace where Sidekick creates editable docs, slides and sheets and incorporates your edits into its next steps.
#AgentsEverywhere
@OpenAI @exaailabs
Had to do a triple take on those claims. Significantly cheaper than fable yet same perf. And fable 5.1 only came out like what, a few weeks ago!?
Pretty fun playing with @Miora_Design's 3D cinematography feature. You can build a whole scene, set the camera angles and plan the camera movement. It’s like having full directorial control, then handing it over to a video model. Here I created a drone shot around a pyramid and gave the agent the 3D reference to generate the finished video.
For this video i used seedance 2.5 and had good results. Will try some of the other models like MiniMax H3. #MioraDesign
You talk it through. Sidekick gets to work.
I built an AI meeting teammate that researches and creates docs, slides & sheets while you talk.
I built this in just a few hours for the @AITinkerers global hackathon.
GPT-Live handles the conversation, while the OpenAI Agents API coordinates research and drafting. Ambiguous AI @ambiguousio is the shared workspace where Sidekick creates editable docs, slides and sheets and incorporates your edits into its next steps.
#AgentsEverywhere
@OpenAI @exaailabs
Meta AI is coming to Hong Kong.
Starting today, with a gradual rollout across WhatsApp, Instagram, Facebook, Threads, Messenger and the Meta AI app.
And it’s powered by Meta’s new Muse Spark model.
Source: LinkedIn
I built an AI focus group that simulates anything
Give it a url and a question and it creates a swarm of agents that argue with each other
I ran it as a focus group for the Grok bot launch. I then sent an update to see what will happen when they make it available to all Cursor users. I could see how agents chatted and convinced each other to change their minds in real time.
it’s a fun demo and kind of mind-blowing what you can now build and deploy in less than two hours!
here's a quick demo video:
I built this in under 2 hours using pages.edgeone.ai/?from=tence… and their Makers Skills in Codex and the EdgeOne CLI to build and deploy it. EdgeOne does the agent runtime, sandboxed browser, models and traces in one project and no need for separate provider accounts or API keys
#tencentcloudedgeone #edgeonemakers
22k GitHub stars as I write this, 90 mins post announcement.
🧩 DeepSeek Harness v0.1 is now available in Developer Preview!
🔹 We’re opening it up to developers building agent harnesses worldwide and open-sourcing the codebase in MIT license.
🔹 Powered by the Cordis meta-framework, DeepSeek Harness is an agent harness built around one core idea: Everything is a plugin. Models, tools, skills, sessions, sandboxes, filesystems, loops, orchestration, and UI are ALL implemented as plugins, and can be mixed, matched, replaced, and extended.
Try it now!
github.com/deepseek-ai/deeps…
Alex Payne retweeted
i'm kinda sick of 5.6 sol now honestly. yes it's a good model, no it doesn't help me ship as much software as it should.
it feels too RL-fried and ends up in all these local minima where it just spins its wheels doing nothing, invents bullshit overly cautious procedures you didn't ask for, doesn't solve your problems, hyperfocuses on tiny details, adds theatrical bullshit like tests/ledgers/hashing/proofs/review loops to "ensure stuff works" (none of it works) and degrades context over long-running threads.
1/3rd of the time it can distil your intent kinda okay and keep it up for a few hours but it will eventually collapse into total slop cannon, no matter how much you steer it. the other 2/3rd of the time it will just misunderstand you and slop cannon.
oh yeah and the limits are cooked and it burns 5x the tokens of 5.5 whilst you don't ship.
bring on GPT 6.....
Kimi K3 has just dropped and it's topping benchmarks. It won’t be the last model that changes the conversation.
One idea I brought back from @RaiseSummit is that the useful question for companies isn’t which model is “best” today.
It’s how much of your product you want tied to any single model or provider.
That’s why @OpenRouter's presentation, and the visibility of companies like @FireworksAI_HQ , @togethercompute and @baseten stood out to me in Paris.
The inference layer can help companies deploy models reliably while retaining the ability to switch.
It's modelmaxxing in practice: test models on your own work, choose the right balance of quality, cost, speed and reliability, and preserve your ability to change.
Baseten were also handing out 'Inference Engineering'. I’ve started it and it's a great read so far!
Moonshot's Kimi K3 is the latest reminder: models will keep changing. Your infrastructure should let you change with them.
This is impressive
It finally finished, here's the final output. Used 60% of my monthly Kimi usage on it
macos27.kimi.page
Alex Payne retweeted
Today we are hosting a GLOBAL Generative UI Hackathon with @AITinkerers
15+ cities, 4 continents. Hong Kong and India are just getting started right now!
@CopilotKit, alongside @GoogleDeepMind, @manufact, @daytonaio and @LangChain
Kicks off with recorded message from co-creators of A2UI, AG-UI and MCP Apps:
@idosal1, @liadyosef, @ataiiam, @zeroasterisk and the organizers
So excited for this!!
Pune AI tinkerers, 12–5 PM is your time to mesh agents with user interfaces.🚀
💎 $25 Gemini credits for every attendee
🏆 Global prizes: Mac Minis, Meta Ray-Bans, $5K GCP credits, LangSmith credits + sponsor swag
pune.aitinkerers.org/p/gener…
Thanks to @AITinkerers @GoogleDeepMind @CopilotKit @manufact @LangChain @daytonaio @aigrantsindia
⏰ Sat 12 PM, demos at 6 — RSVP
Alex Payne retweeted
Meet Mandarin Quest.
A voice-first Mandarin learning game where CopilotKit directs the quest UI and A2UI renders the coach in real time.
Speak naturally, complete night-market missions, and watch mistakes turn into corrections, drills, and reviews.
Not a chatbot. Generative UI for language learning.
#genUIHackathon @AITinkerers @googledevs @CopilotKit
Meet Mandarin Quest.
A voice-first Mandarin learning game where CopilotKit directs the quest UI and A2UI renders the coach in real time.
Speak naturally, complete night-market missions, and watch mistakes turn into corrections, drills, and reviews.
Not a chatbot. Generative UI for language learning.
#genUIHackathon @AITinkerers @googledevs @CopilotKit
New Claude code tasks sounds great, particularly this feature:
“When one session updates a Task, that is broadcasted to all sessions currently working on the same Task List.”
