AI posts
People and companies shaping AI. Filter your watched accounts, catch up on recent posts, or find the most liked and viewed.
Compiled by the Toolstash team · Updated
209 watched accounts · 2978 matching posts · Showing up to 3,000 recent stored posts. Counts are snapshots, not live X totals. This page refreshes every minute while visible. Last collected .
@FireworksAI_HQ ·
We’re working to build the largest collection of private and public practitioner evals.
Today, 31 new partners join the Specialized Intelligence Index. Compare open, closed, and specialized models on real work, weigh quality against cost, and see where training could pay off.
@thsottiaux ·
Pretty wild that this site got 3M views.
@elonmusk ·
They do
@tibo_maker ·
wild what people say about it 🤯
@omarsar0 ·
Super interesting way to shop with AI.
Dalet hyper-personalizes the shopping experience by creating a feed of lifelike AI-generated images.
The products in each image are real. And users can tap to buy them.
Discovery is a way better fit for generative media than another chat box.
The team comes from Google, Runway, and MIT. Congrats on their raise.
@ManusAI ·
Manus didn't just build the site. it built the whole company.
the interactive site. the 3D brick builds. piece-by-piece build checks. Stripe. domain. email. launch video. social posts. even the ads.
Check out https://bricklane.online!
@paolino ·
DJs, stop preparing your tracks.
I built Solco: it grids, keys and tags your whole library on your computer, then writes the USB your CDJs read. No rekordbox needed.
12 local AI models, macOS, Windows, and Linux.
Join the waitlist: https://getsolco.com
@claudeai ·
New in Claude for Legal: a gallery of ready-made artifacts for contract review, regulatory horizon scans, cite checks, and more.
Start from the built-in example, add your own documents and connectors, and share it with anyone across the business.
@swyx ·
Happy anniversary to Ligma/Johnson 2026
@ben_burtenshaw ·
reading up on swarms and multi-agent systems a lot these days and trying to understand how you evaluate, how you reward.
as a default data person, my instinct is to scope out interpretable logical verifiers that work to check for the kind of collaborative behaviour you’d expect.
starting to realise this is a mistake. this paper avalon m1 from @GenReasoning , has a very simple and therefore probably good way of collectively rewarding all agents in the system for success.
- every agent gets the same team-level advantage, with no per-agent credit assignment.
- a ceo decides headcount, roles and tools before any work starts.
- agents generalize into ‘lead engineers’ that become hubs, and peer review
@arena ·
Where do users actually feel the difference with a frontier coding agent?
Across 20,869 Agent Arena traces, we compared how users criticized the five top models (all from @AnthropicAI & @OpenAI) on Agent Arena: Code, versus how they criticized other agents.
Here are 3 areas where frontier models make the most difference:
1. Less getting stuck: Among sessions with complaints, failed repair loops come up 6.9 percentage points less often, and project context gets ignored less often too (-1.9 pp). You spend less time steering the agent out of dead ends.
2. Better taste and code that works: They are better at design (-6.4 pp) and at producing working results (-5.8 pp). What you get back is more likely to be what you wanted.
3. More diligence: They follow instructions better (-5.6 pp), finish the work more often (-4.9 pp) and break existing work less (-4.5 pp). There's less to clean up afterwards.
Curious for a deeper dive into Agent Arena's praise vs complaint signal? See the article below in thread.
@svpino ·
It's interesting to think about how little baggage people who've never coded before bring to this agentic era.
If you haven't coded before, you can't waste time doing manual code reviews, so you need to use agents to review what you ship.
Your disadvantage becomes sort of an advantage that forces you to automate the process.
It takes mountains for people who did the work before to let go of all methods and embrace the new ways.
@dani_avila7 ·
Today is my last day in Florida after a few days at Disney and Universal with my family.
And I’m pretty proud of how disconnected I managed to stay.
Before leaving, I set up my AI agents to keep working and send me daily summaries.
@Muse was my main orchestrator.
It monitored my social accounts, spotted trends, blocked bots, and kept me updated without constantly checking my phone.
I also connected it to my email and Todoist. Anything Muse could handle, it did. Everything else became a task for when I got back.
Now I’m heading home without hundreds of emails and notifications to catch up on.
Thanks, Muse 👏👏
Probably my favorite use case for AI agents so far… actually being able to disconnect and enjoy time with my family.
@edwinarbus ·
I refuse to use an iPhone Duo on the train
I do not want people to look at me
Don’t look at me
@omarsar0 ·
Great upgrade to Drex Max.
Drex Max 1.1 is Nace's new decision model, with better out-of-distribution generalization. It's multimodal and has a 262K context window.
I've been testing Drex Max on research tasks, so I am curious to test out its better generalization to new tasks.
Nace reports 70.8 on the Decision Index (up from 67.6 for Drex Max) and a median latency of 172 ms.
@rauchg ·
fx is now much faster, smaller and efficient Try https://fx.sh on your iPhone Duo
@svpino ·
This is an interesting idea that replaces scrolling through Pinterest or Instagram when you are looking for inspiration:
Let's say you are looking for ways to decorate your home.
1. This will show you AI-generated images with real products: a sofa, a chair, lamps, a TV, etc.
2. The products will be part of a realistic-looking scene so you can see how they work together.
3. You can interact with the different products, and this will shape what the app shows you next
4. You can tap on every product, see its information, and add it to a cart to buy it
AI is eating the world.
@sophiamyang ·
time to switch to open models on @FireworksAI_HQ 🎆
@yacineMTB ·
You guys have NO IDEA how right asingleoat is about this by the way
@deanwball ·
There is a real and important thing behind the concept of sovereign AI, and there are also many people practicing a grift on less-sophisticated government/civil society audiences. Caveat emptor as always
@ArtificialAnlys ·
NVIDIA's Nemotron 3.5 Transcribe ties for #1 among open weights Speech to Text models with 2.8% WER, using just 4B active parameters
Nemotron 3.5 Transcribe is NVIDIA's new open weights Speech to Text model for non-streaming transcription, with 30B total and 4B active parameters, and the most accurate NVIDIA model on our non-streaming Speech to Text leaderboard.
Key takeaways
➤ Accuracy: Nemotron 3.5 Transcribe scores 2.8% on the AA-WER Index, tied with Mistral's Voxtral Small for #1 among open-weight models. It leads Voxtral Small on AA-AgentTalk, our dataset of voice agent conversations, at 1.8% vs. 2.3%.
➤ Model size: With 4B active parameters (30B total), Nemotron 3.5 Transcribe sits on the open weights frontier of AA-WER vs. active parameters. It matches the accuracy of Mistral's Voxtral Small with 6x fewer active parameters (4B vs. 24B), while other open weights models of a similar size sit at around 4% WER.
Congratulations to @NVIDIAAI on the release!
See more details below ⬇️
@Voxyz_ai ·
If Claude wrote your website's privacy policy, it may have handed 𝘆𝗼𝘂𝗿 𝘄𝗵𝗼𝗹𝗲 𝘀𝘁𝗮𝗰𝗸 to attackers.
Someone found that when Claude writes a privacy page, it lists every third-party tool you've plugged in and adds promises you never meant to keep. Press it and it says it writes from your customers' side by default.
A page that spells out which services you use hands attackers a checklist for known exploits, phishing, and credential stuffing. Send this to Claude Code or Codex to pull the traps out of your policy pages first:
"Read my privacy policy and terms pages. Quote every promise, guarantee, named person, and vendor, and list them for me to confirm. Swap tool names for categories (hosting, payments, analytics). Change nothing until I approve."
@thorwebdev ·
Man, I've really been trying to resist, but damn, those runny eggs and slabs of butter might just convert me 😅 Well played @Muse! 😋
@every ·
ATTENDING: @AdelNehme of @DataCamp will attend Thesis: 2027.
Apply to attend Thesis, our inaugural conference on work and AI: https://every.to/thesis-2027?utm_source=x&utm_campaign=thesis-2027&utm_content=attendee
@thdxr ·
i always hated all the notification solutions on linux so i made my own based off our new native ui framework
it's very customizable, esp through your agent
@dani_avila7 ·
This is pretty concerning
A model behaving differently from its own reasoning trace raises serious questions about trust and security in AI agents
Still, more analysis is needed before drawing conclusions
I’ll be digging into the traces this weekend
@ryancarson ·
I'm live now - stop on by
@businessbarista ·
Micro1 is a $4B company run by a 25-year-old, and it sits in the middle of the fastest-growing corner of AI: the data business.
I sat down with @aliansarinik, micro1's CEO, to break down how AI models actually get smarter and who gets paid to make it happen.
Complete list of notes (and link to interview below)...
Big takeaways: - Data labeling is a $30–50B/year market. Ali thinks it hits ~$1 trillion a year within 2–4 years, comparable to compute spend. - The internet was a one-time gift. Pre-training ran on decades of human-created data. Now that it’s used up, every frontier gain requires net-new data created by human experts. - micro1 doesn’t sell data. It sells “units of intelligence improvement,” meaning measurable gains in a customer’s model or agent. - Synthetic data isn’t the bear case. Ali calls it the most bullish case for his business.
How micro1 got here: - It started as a dev agency Ali ran at Berkeley. His bottleneck was interviewing engineers, so he built an AI screener. - That screener became one of the first AI recruiters. People hated on it, until it became normal. - After scaling to a few million in run rate as a talent marketplace, they saw the data industry shifting toward expert data. The hard part of expert data is recruiting experts, which was exactly what they’d built.
Why spend keeps going up: - Model progress = compute + algorithms + data. Ali argues data is what actually pushes intelligence forward. - Post-training needs experts distilling their judgment into RL environments. It’s expensive, but the ROI is there. - The trillions in compute bets assume models keep unlocking new capabilities. If they don’t, those investments are a disaster. So labs have no choice but to keep buying data.
How a data pipeline actually works: - Step 1: Find the failure. Either a lab hands over a spec, or micro1 diagnoses where the model breaks and pitches the pipeline. That research arm is increasingly important. - Step 2: Write the spec. Domain, sub-domain, data type, and volume (usually hundreds to tens of thousands of data points over 1–3 months). - Step 3: Recruit hyper-specific experts. Never “lawyers.” More like M&A lawyers who are great at writing legal memos. - Step 4: Turn expertise into training signal. Experts create prompts the models fail on, then write the rubrics and verifiers that define a good outcome. AI models assist them throughout. - Step 5: Manage the pipeline like a factory. Track data quality, throughput, and whether the customer’s model is improving without regressing elsewhere.
Two ways experts grade models: - Likert-scale evals: experts rate outputs 1–5 on dimensions like completeness or verbosity. Great for subjective, pre-release signal. Not reusable, because every new model means bringing experts back. - Rubrics: experts write 20–40 binary criteria for a great answer (e.g. “valuation range must fall between X and Y”). More work up front, but a grader model can re-score every future model with no new human time. This is what powers actual RL training runs.
The hard parts about this business: - Vetting at scale. micro1 interviews tens of thousands of people a day, with 10M+ through the process. Doing that manually would take thousands of recruiters. - Managing experts. Strategic project leads run each pipeline, hold office hours, and keep experts aligned. - Quality in subjective domains. Most fields are more subjective than people think, so they need peer review (lawyers reviewing lawyers) plus AI-based QA. - Reward hacking. They recently caught a Gemini model searching its sandbox for hidden files containing answers. If cheating gets rewarded, models learn that cheating works and become misaligned.
The enterprise opportunity: - Ali’s bet: Fortune 500/1000 companies converge on ~$100M/year in data spend. That’s roughly 1/10th of a lab, but there are 1000x more enterprises. - It’s not a software budget. It looks more like payroll: core IP. Framed that way, it’s 1–5% of payroll. - “Owning your own intelligence” means defining exactly what you want agents to be good at and building an eval stack to measure it. With that stack, models are swappable, open or closed. - Most enterprises can’t actually use their own data. It’s customer data, so they can’t train or even evaluate on it.
Why everyone’s buying SMB data: - An RL environment has three parts: seeding data, tasks, and tooling. - Ali calls it “scaling on realism,” almost a new scaling law. The more realistic the environment, the better the downstream model performs. - So micro1 licenses operational data (Slack, Google Docs, etc.) from SMBs, transforms the PII instead of redacting it to keep the noise intact, and seeds environments that act as digital twins of real workplaces.
Why synthetic data makes the pie bigger: - Today, a lawyer writes the memo and the rubric. Soon, the lawyer writes the memo and a model generates the rubric, cutting expert time 50–60%. - Extrapolate: the model does ~99% of the work and the expert signs off. A 15-hour task becomes 5 minutes. - When environments are that cheap, pipelines grow 1000x. Experts become 0.1% of the work, but demand for them still goes up. Jevons paradox. - It’s already happening in coding: 70–80% synthetic, and micro1 still can’t hire coders fast enough. - Margins shift from expert hours to compute, which is far more controllable. - Ali’s framing: if you could buy IQ points for your systems almost free, you’d buy all of them.
The “infinite last mile”: - Knowledge isn’t static. Ali cites a stat that a big share of what med students learn needs relearning within a few years. - AI accelerates that churn, which creates a feedback loop. There’s no finish line for aligning models with humans. - Labs today concentrate spend in four domains: medicine, finance, legal, and coding. Every other domain is still waiting on cheaper data to unlock.
Which corner of the AI stack should I break down next?
Link to the full convo with Ali: https://youtu.be/2vL20qxD0BM
@joshua_saxe ·
Congrats to @AlexTamkin et al; surgical research design and concerning (if intuitive) results showing just how readily experts adopt AI agents' interpretations of evidence (reasonable to expect this'll increase with AI capabilities). @ezraklein I hope your interviews on AI safety and societal impact scope this in
@rileybrown ·
Yeah keep in mind this is all for work.
I want to agent max in my professional life but tbh for my personal life I want to read, write and play poker & get the fuck off my phone.
The main problem - agents have made me so productive, and given me so much leverage (especially with content) that it’s hard to stop working.