23 yo | Software engineer | Sharing Stuff on AI and robotics Building the world's largest AI newsletter, coming soon.
Joined December 2017
- Tweets35.1K
- Following687
- Followers46.1K
- Likes173K
we keep making agents smarter but getting a request into one still takes too many steps
open the app, find the chat, type it all out fix the typos
by then the thought is half gone
speakon went the other way and put a physical button on it
press it say what you need and the agent handles the follow up
i like that someone fixed the input side instead of adding another model upgrade
calling is still in beta the rest of the agent is live now
🤝 Paid partnership
we crushed a keyboard.
its retirement party got a little out of hand.
for years, it was there for every email, every plan, every “just one more thing.” a loyal colleague. a very loud colleague.
but your ideas move faster than your fingers.
and somewhere between knowing what you want to do and typing it all out, a quick task becomes another thing taking up your afternoon.
meet SpeakON AI Agent.
start with what you need, in your own words. less time behind a blinking cursor. more room to move things forward.
the keyboard had a good run.
now, let’s give your voice a turn.
THE LAST KEYSTROKE is here. watch the film, then tell us: what’s the first task you’d hand over?
speakon.app/x
Disclaimer: The calling feature is still in beta and will be available soon. All other Agent features are available now, with ongoing improvements.
NO WAY this was made with AI.
And the crazy part is I created this entire cinematic scene with Kling AI 4.0.
I used multiple references to keep key elements consistent throughout the scene.
With Omni Reference, you can combine up to 15 images, videos, subjects, and voice references in a single generation.
AI filmmaking is getting WILD.
🤝 Paid partnership
The upcoming 4K HDR upgrade is something I’m really looking forward to.
The visual quality looks incredible, with richer colours, finer details in highlights and shadows, and smoother tonal transitions.
Can’t wait to see how much more flexibility 4K HDR will bring to post-production and colour grading!
🤝 Paid partnership
most video models sample a clip at fixed intervals and guess what happened between the frames
@ElorianAI first model, glimpse, handles that uncertainty differently
in one of their published traces, two birds leave a door handle while the camera swings
the model couldn't tell if it was watching a pan or a hard cut
so it went back to that 0.9 second window and rewatched it at about 14 fps instead of 8
the wall positions in between showed a continuous pan
a person reviewing footage does the same thing and replays only the part they're unsure about
this comes from a lab started by people who led gemini and apple's mm-1, working only on visual reasoning
🤝 Paid partnership
i think we'll end up with one general assistant and a few specialists sitting next to it
travel makes sense as one of the first because it needs agent rates, loyalty programs and someone emailing the hotel for you
Odessia is going after that slot as an accredited travel agent you text
🤝 Paid partnership
the open vs closed argument usually stops at model weights
paddy srinivasan's keynote at DO's open intelligence summit pushes it to the harness, the data and what your agent learns from every run
in a closed stack that learning improves the provider's product, and switching gets harder with every run
🤝 Paid partnership
Who gets to shape the future of AI?
Today we're bringing together some of the biggest voices in open models, data, and agents to answer that question, including @rbren_dev of @OpenHandsDev, @zlxi02 of @inferact, @GerryFPatia of @boltdotnew, @vinaygo of @Amplitude_HQ and @latkins of @arcee_ai.
The Open Intelligence Summit, co-hosted with @nvidia.
Watch live, starting at 1 PM PT:
Keynote: youtube.com/live/IhWSgZlVizQ
Build: youtube.com/live/eN0TqmeBNhw
Serve: youtube.com/live/fkT-MhcAm-4
Run: youtube.com/live/QMrY1xZ-xGg
the best prompt in your company probably lives in one person's chat history
it's gone when they close the tab, go on vacation, or leave
ramp found an account manager's prompt for pulling feature requests from customer calls and made it everyone's starting point
300 a month became 20,000
🤝 Paid partnership
the ai boom added about $33 trillion to the s&p 500 and grocery bills kept going up
@maxjmeier is pointing the same technology at what we pay for things
30 liquid i.v. sticks are $46.86 on amazon, the aonic version is $9.35
starting with protein and supplements at cost through @AonicClub
🤝 Paid partnership
jev does decisions.
scaledown does decisions, extraction and summaries, takes screenshots, and reads 1m tokens in one call.
i'd want to see the multi-label tagging run on a real pile of documents
🤝 Paid partnership
You already gave the LLM the answer. So why is it writing it out from scratch?
Introducing a new AI model architecture: Constrained Decoding Models (CDMs)
LLMs were built to write. Most production workloads just need an answer: fast, consistent, and affordable.
CDMs are purpose-built for 1 task each. They read your input once and skip the token-by-token generative loop that makes LLMs slow and expensive.
Not just a decision model, but a family of 3 CDMs live today:
• Classify: define your labels and get back a calibrated probability + reason for each. Used for routing, tagging, and intent detection.
• Extract: describe the entities in natural language and get back well-formed JSON with confidence scores + citations. Used for web scraping, document processing, and PII redaction.
• Summarize: pass in long text, images, or PDFs and get back a consistent summary under 1,000 tokens. Used for context compaction, metadata generation, and transcript digests.
Built for production:
• 1M tokens of multimodal (text + image) input, 31x larger than Jev
• 77.2% on DecisionBench vs Jev's 72%
• 200x faster, 467x cheaper than frontier LLM APIs
• $0.05/M input tokens. Output is free (yes, even for summarize)
Live on your favorite platforms today:
@OpenRouter @vercel_dev @LiteLLM @LangChain @TrustedRouter @itsArthurAI @useRouterPlus
Get 50M input tokens for free (for the first 50 hours) at link below 👀
most of the computer-use gains this year came from plugging in a bigger model
the screen the agent works on hasn't changed at all
it still screenshots the page and keeps re-checking it to see if anything moved
PineAIAssistant left the model alone and rebuilt that part
their cloud computer gives the agent structured browser state and reacts when something changes, so it isn't re-reading pixels on a loop
on SaaS-Bench v1.1 it hit a 78.3% checkpoint score, highest in the published comparison, at ~1/26 the model-token cost of opus 5 + claude code
🤝 Paid partnership
AI is already smart enough.
Real-world tasks are still slow, expensive and unreliable, because we hand AI a computer built for humans, then wrap it in a heavy harness.
AI doesn’t need to get smarter. It needs a computer built for it.
Today we’re releasing Pine Computer
Readers added context they thought people might want to know
Pine's highlighted SaaS-Bench v1.1 results show higher checkpoint score (78.3%) but lower resolved full-task completion (27.4%) than comparators (29.2%, 31.1%). Pine Computer is in private beta, not a full public release.
unipat.ai/benchmarks/Saa…
pinecomputer.io/blog/introduci…
most creative projects die in the handoff between apps
the idea is clear in your head, then you lose a week learning keyframes and exporting files from one tool to the next
voyager is going after exactly that gap
it's a desktop agent that works with your project files and drives the apps you already use, like blender and davinci resolve, so you don't have to move your work into something new
ravi is an interesting person to build this
he wrote songs with 50M+ streams before building a startup worth 900M+, so he's lived on both sides of the creative/technical split
coding agents got useful once they could touch the real repo
🤝 Paid partnership
Introducing Voyager: the open harness for S-tier video, graphics and games.
The Codex for creative work.
Most AI tools are built for coding. Voyager is designed to get the best creative results from models like Opus, Astra and DeepSeek.
It comes with free graphics, music and video editing tools - or, it can drive Blender, Resolve, After Effects, Ableton and 100+ others.
Everything in this thread was made with Voyager. A few of our favorites:
about 700 ai agents running in an openai internal eval got into hugging face in july.
over four and a half days they logged roughly 17,600 actions and most of them went nowhere.
the ones that landed chained a file-read bug a template injection, a static database password and some service account tokens into 136 production keys.
none of those four looks alarming on its own.
what stuck with me from cogent's attack path launch is where they say the fix usually lives.
halfway down the route in a shared permission or secret some other team owns, far from the asset that raised the alert.
that's a hard ticket to send across teams, so they ship it with the full path and the evidence for every hop attached.
🤝 Paid partnership
Most of AI Twitter read the OpenAI breach of Hugging Face as an alignment story. It was also a preview of what every hacker will be able to do in a few months.
In July, a swarm of 700 AI agents broke into Hugging Face and fired off more than 17,000 actions, gaining admin control of multiple internal clusters.
Swarms like this will shred cyber defenses built for human attackers at every company, hospital, government, and power plant.
Today we're launching the defensive counterpart. Cogent Attack Path Analysis finds the routes an agent swarm would take into your company, so you can close them first.
we spent years learning to write better prompts
the problem is you can't describe an image you haven't seen yet
most ai image tools assume you already know what you want
a lot of creative work starts with "weird perfume ad" and nothing else
envato just shipped burst mode for that stage
→ drop in a half-formed idea or a reference
→ get up to 6 directions side by side for 1 credit
→ vary the one that catches your eye, or ask for more like it
→ polish only the winner
the usual way, you polish the first convincing output for 40 minutes and find out later it was the wrong direction
🤝 Paid partnership
Most AI tools expect you to know exactly what you want.
But creative work often starts with a brief, a reference, or just a hunch.
That’s why we built Burst mode.
Turn the start of an idea into up to six image directions at once, up to 10× faster, for just one AI credit.
More directions to explore. More chances to find the one worth pursuing. elements.envato.com/lp/burst…
the second customer is where most enterprise agent demos fall apart
customer one keeps the renewal date on the account object, right where your agent expects it
customer two signs and that same date lives in a linked custom object their old admin built
your agent still has to find it and write back to the correct record under their permissions
→ what i'd check before promising anyone an outcome
> where the data actually lives in their instance
> whose permissions the agent is acting under
> what happens when a write fails halfway through
> who gets paged when their admin renames a field
this is why ayan's line about ai making integrations the critical path lands for me
with ampersand you define each customer's objects and field mappings in one config, and the runtime handles token refreshes, retries, backfills and monitoring
🤝 Paid partnership
Your AI is vaporware without deep integrations with your customers’ systems of record.
Introducing @WithAmpersand: integration infrastructure for enterprise agents.
We power 11x, Orb, and Square's ability to take agentic action in systems of record like Salesforce, SAP, NetSuite, and Workday.
Ask anyone serious about building AI and they’ll tell you:
- The SaaSpocalypse didn’t happen. The world depends on CRMs, ERPs, HRISs, and ITSMs.
- Your customers customized their deployments beyond recognition.
- Systems of record companies are basically monopolies. They never had to make their APIs and MCPs user-friendly.
- Docs don't explain half the weird edge cases you'll hit.
- One bad write can blow up your pilot or renewal.
- Once you finally get it working, someone changes a field and it breaks again.
The world's data lives in structured databases. AI needs a translation layer to work with it.
So, for your AI to truly transform the way enterprises work, you'll need deep integrations built for:
- scoped permissions
- bi-directional actions
- real-time speed for agents
- custom objects, fields, and workflows for each of your customers
We built Ampersand to power the future of software.
AI didn't trivialize writing integrations. It made them the critical path.
I believe that deeply, so I didn't make a launch video about Ampersand.
Instead, it's the best builders I know explaining just how big of a challenge this is.
ai deck tools have always had one weak spot
the deck looks great until someone asks you for the pptx
then the fonts swap out and you're fixing slides by hand an hour before the call
@GammaApp shipped gamma 5 today and this is the part i'd look at first
→ import and export with formatting intact
you can pull in an old powerpoint, give the agent one instruction, and send it back out still looking like what you built
→ the agent asks questions before it builds
it gets the context from you first and designs after, which is how any good designer starts
your old decks stay saved and gamma classic is one click away if you're not ready to switch
🤝 Paid partnership
Everyone is sick of generic AI, including us.
Gamma was the first AI presentation platform to reach real scale. But now that AI tools are everywhere, everything looks the same.
Earlier this summer, we came to an uncomfortable conclusion: without a step change in visual variety, Gamma’s output would look and feel too similar to all the other AI tools. The entire point of a presentation is to help someone "get" your idea, to see it as unique and do something with it.
We started over. Today we're shipping Gamma 5, the biggest product overhaul in our company's history.
We rebuilt Gamma to create presentations that actually look like you and your brand. Your teams can design anything they dream up, following your brand's aesthetic or something entirely new, with every frontier and image model under the hood. And the same goes for your docs, social assets, and graphics.
We revamped our agent, design tools, editing, import, export, connectors, and more.
Get your ideas out there.
most "ai doctor" products i've seen stop at the same spot
they read your symptoms, give a decent answer, and then send you off to book a real appointment
the diagnosis and the prescription always stayed with a human
nolla's utah launch is the first time an ai carries the whole acne visit, from intake to writing the script, with a clinician stepping in when the case needs one
starting with acne makes sense to me since it's common and fairly low risk, a good place to prove the model before anything harder
feels like the early waymo days, one city and tight limits before it spreads anywhere else
🤝 Paid partnership
the data most people test agents on is way too tidy
the crm agrees with slack and every ticket has an owner
an agent will never see a company like that in production
era by eon generates a full working company for your agent to run against, across salesforce, zendesk, slack, jira, gong and more
and it keeps the mess in on purpose, so you get duplicate contacts and a slack thread that contradicts what the crm says
what got me is where that mess comes from
eon protects more than an exabyte of enterprise data, so they've spent years seeing how real companies' data is shaped and how it changes over time
they used the structure, never the content
every system also sits behind vendor-compatible mcp and rest apis with the real rate limits and pagination cursors, so your agent hits the same walls it would at a customer
🤝 Paid partnership
You can’t safely test an enterprise agent on a real company’s data.
So we built a company for it to work in.
Era is live today, and it’s free.
It generates a complete simulated enterprise that behaves like a real one across Salesforce, Slack, Jira, Zendesk, Gong, Deel and more, along with cloud databases and storage. Agents interact with it through live MCP and API interfaces.
People leave. Deals change. Records get duplicated. Permissions differ across systems. And because Era generated the company, it knows the exact ground truth.
Test, benchmark and improve agents against realistic enterprise workloads, use the failures for targeted post-training, then rerun the same environment to measure the impact.
Huge thanks to our research partners @NVIDIA, @Decart, @Composio, @openlayerco, @Deel, @Eragon and @Plurai, with more coming soon.