https://nitter.gq/t.co/GS8VHacpw5 by day, https://nitter.gq/t.co/pwsQWLthAD by night. ex-microsoft, ex-HP, ex-Monster. My opinions are my opinions :)
50.077662,14.416902
Joined May 2007
- Tweets4K
- Following1.6K
- Followers383
- Likes5.4K
Nikola Malovic retweeted
now that apple put xcode inside claude code, here’s how i’d find, build and grow an iOS app with AI agents:
find it
- open appfigures or sensor tower and sort utilities by revenue
- filter for apps with a rating under 3.5
- bonus points if the top app hasn't shipped an update in 6+ months
- check the keywords they rank for in appfigures. low difficulty + high revenue is the sweet spot
- scroll the "you might also like" row on the top app's page. every app there is another market to check
- or find a startup idea on ideabrowser.com based on social sentiment
validate it
- use facundoolano/app-store-scraper to pull every review for the top 20
- give them to an agent: "cluster the 1 and 2 star reviews into the top 5 complaints and quote the users"
- pick the app where the complaints are about the product, not the price
- if people complain about the price but keep paying, that's demand. if they complain about bugs, that's your opening
- search reddit for "alternative to [app]". every thread is a list of people waiting for you
build it
- claude code + apple's new xcode plugin to build, preview and fix
- cameroncooke/XcodeBuildMCP so the agent can build, test and run the simulator on its own
- joshuayoes/ios-simulator-mcp so it can tap, swipe and screenshot like a user
- write a skill with the top 5 complaints so every feature gets checked against them
- apple's foundation models for on-device ai, so your inference cost is $0
- app intents so siri and personal agents can use your app too
- add an "import from [competitor]" button. make switching take 10 seconds
test it
- maestro tests every flow the reviewers complained about
- the agent runs them on every screen size before each build
- if the competitor's worst flow takes 6 taps, yours takes 2
- sentry for crashes, telemetrydeck for analytics
- testflight it to 20 people from those reddit threads before launch
monetize it
- revenuecat for subscriptions and mrr
- superwall to test paywalls without shipping a new build
- paywall right after the first win, never on launch
- annual as the default with a 3-day trial, weekly as the anchor
- turn on win-back offers for everyone who cancels
rank it
- your screenshots answer each complaint directly. "no ads." "works offline." "cancel anytime."
- fastlane snapshot + frameit regenerates every screenshot automatically
- a/b test them with app store connect's product page optimization
- have the agent localize your listing into 10 languages. most competitors only rank in english
- ask for a review right after a win with apple's built-in review prompt, never after an error
grow it
- find 10 creators in your niche on tiktok one. small monthly base + a bonus for every 100k views
- give each one a fresh tiktok account that only posts about your niche
- apify's tiktok scraper pulls the top 100 hooks in your niche every week. that's your hook bank
- every creator tests 3 new hooks a day in capcut. anything that loses people in the first 3 seconds gets cut
- arcads or heygen for ai ugc when you need 50 variations of a winner
- postiz to schedule everything across tiktok, reels and shorts
- faceless slideshow accounts posting 3x a day: problem, screenshot, result
- when an organic video takes off, put money behind it with tiktok spark ads instead of making new ads
- a unique apple offer code for every creator so you know exactly who drove each subscriber
- a quiz landing page on framer that ends in "download the app." web-to-app funnels convert better than cold traffic
- appsflyer to track which video, creator and ad actually turned into revenue
- apple ads on the competitor's name
i could do a whole episode on @startupideaspod if people are interested. i will never gatekeep anything im learning in real-time.
welcome to the golden age of ios apps, because now anyone can build one
Apple just put Xcode inside Claude Code.
It's an official Apple plugin, now in Claude Code's plugin directory:
• Build, run and test your app
• Render SwiftUI previews
• Drive simulators and real iPhones
• Read build logs and errors
• Search Apple's developer docs
Claude can now build your iPhone app, see what it looks like, and fix it without you opening Xcode.
Install in Claude Code:
/plugin install xcode@claude-plugins-official
Needs the latest Xcode on your Mac.
Nikola Malovic retweeted
running a company nowadays feels kinda dystopian
my team is 50+ people in 12 countries, but i talk to my agents more than my human teammates
- voice note to grokbot
- mattpocock/skills /grill-me so the agent interrogates the idea before anything gets built
- spec drops into linear through the linear mcp
- exa + firecrawl agent reads every competitor site, pricing page and changelog
- ideabrowser.com to find ideas and trends
- apify pulls g2, reddit and app store reviews
- prompt: "find the 10 complaints that show up most and quote the users"
- context7 so the agent reads the latest docs before it touches any api
the design
- claude code + paper mcp
- 4 versions of every screen before a single line of code
- pick pieces from 2, it merges them. diverge then converge
- frontend-design skill from anthropic/skills + our own design system skill
- cleanshot screenshots of any ui i like go straight in with "make it feel like this"
the build
- pingdotgg/t3code as the control surface for every agent, from my phone too
- stablyai/orca runs a fleet of agents in parallel, each in its own git worktree
- graphify maps the whole codebase into a knowledge graph so agents navigate instead of guess
- headroom compresses tool outputs before they hit the model, so tokens go 20-95% further
- lsp plugins so the agent sees types and errors like an ide does
- security guidance hook flags risky edits before they land
- supabase mcp for the database, stripe for billing
- github mcp opens the pr, vercel spins up a preview
- loops run overnight, i check what shipped over coffee
the user testing
- lightpanda as the headless browser, way lighter than chromium so you can run hundreds at once
- agents click every flow on the preview like a first-time user
- one persona is impatient, one is confused, one is on a phone
- playwright + chrome devtools mcp for regression tests
the bugs
- sentry mcp catches the error
- agent reads the stack trace and opens a fix pr in under 10 minutes
- human just approves
the data
- posthog mcp cron at 7am on yesterday's funnel
- any step that drops more than 10%, the agent watches 20 session replays and tells me why
- stripe mcp pulls every cancellation reason into one doc every monday
the agents' tools
- orgo so every agent gets its own computer
- agentmail so every agent gets its own inbox
- agentphone so i can call them from the car
- composio connect-apps for every other connector
- akitaonrails/ai-memory, a git-backed markdown wiki so they remember what we decided last week
the skills
- wshobson/agents so skills work across claude code, codex and opencode
- my writing voice
- our design system
- how we write specs
- skill creator so the agents write new skills themselves
some helpful tricks
- one agent per platform. the x agent never touches email
- cheap model for triage, opus for anything that ships
- every output gets scored 1-10 by an llm judge. under 8 goes back
- every agent writes a 5-line log at night so i can read the whole company in 2 minutes
- every agent mistake becomes a new skill, so it only happens once
the team
- humans get version 5, not my voice note which they are happy about lol
- their job is creative feedback, relationships and telling the agents what they got wrong
a human pm reviews the funnel once a week. this reviews it every morning
running a company is now voice notes and inference
Nikola Malovic retweeted
Launching RNAFoundry.com – we manufacture personalized cancer vaccines.
Nikola Malovic retweeted
this might be my last normal tweet before biotech goes vertical
i'm not a whistleblower. I don't work at Lilly, Novo or any of the big labs
but there is a massive shift happening in human biology right now.
it's sitting in press releases and almost nobody reads them
the last 6 months alone:
- a once-daily pill roughly doubled median survival in metastatic pancreatic cancer (13.2 vs 6.6 months)
- a one-time gene editing infusion cut hereditary angioedema attacks 87%, and 62% of patients stayed attack-free and off all other prevention for 6 months
- a one-time gene edit for cholesterol dropped LDL up to 62%, and it held for over a year
- a personalized mrna vaccine beat keytruda alone in a 1,137 patient melanoma phase 3
- retatrutide hit ~28% weight loss in a phase 3
- an ai-designed drug entered phase 3
- a pill that regrows hair passed its second phase 3
- an immunotherapy that could let some rectal cancer patients skip surgery, chemo and radiation just got fda approved
- a man lived 271 days off dialysis on a pig kidney
- human cells were caught passing chromosome-sized chunks of DNA to their neighbors, and it changes how the neighbors behave
we are on the verge of an era where the human body is a software problem and the update cycle gets shorter everyday
stick around, it's going to get weird
bio/acc
Nikola Malovic retweeted
Ben Affleck reveals he writes Python, understands convolutional neural networks, worked extensively with GPUs, and used his celebrity status to get private looks at Google and OpenAI’s video models
“I’ve always been kind of into computers since I was young. Then, when film started to move from analog film to digital, I became more interested in that aspect of it. The visual-effects workflow for many years has included machine learning, so I can write pretty shitty Python scripts and stuff like that.
“With convolutional neural networks, which were the precursors to what the transformer can do, which is much more computation simultaneously, you would do things like look at what’s called a tensor. That’s the numerical translation of a visual image in numbers, like the batch number, the frame number and the red, green and blue values of each pixel in each frame. It’s just that simple. That numeric is called a tensor.
“You’d use a convolutional neural network to identify patterns that reveal what’s called edge detection or feature extraction, which is identifying patterns well enough to know, this is where the window ledge is, so we can more easily take the green-screen image out and replace it with something.
“That was familiar to me early on because, prior to Artists Equity, I had a small visual-effects company. I’ve worked with GPUs a lot too. The visual-effects guys said, ‘Hey, you should see. There are a couple: Google and this other company, OpenAI, are doing really interesting stuff with transformers in video.’
“I’ve learned that I can actually just call up and go, ‘Hey, it’s Ben Affleck. Can I come see what you’re doing?’ Sometimes people say yes, to my astonishment.”
$ANET Andy Bechtolsheim says AI has made optics demand about 10 times bigger in five years and the industry is only at the "very beginning"
"So I guess I don't need to tell you that AI has been driving this incredible increase in demand for high-speed optics, which is probably now 10 times bigger than it used to be five years ago..."
"And what I want to talk to you today is that we're not at the end of this journey, but rather the very beginning. There's easily another order of magnitude increase in bits needed for the next generation kind of data centers."
"So what I want to talk about first is what's happening at the data center level, and how we as an industry have to get together to solve these problems of this very rapid and high demand growth as fast as possible, since AI can't wait."
______
For full set of takeaways to Andy's AI Datacenter and Optics lecture: firesidealpha.substack.com/p…
Nikola Malovic retweeted
This is actually insane. This uses @typesafeai Jev model, as a plugin in Claude to review all the un-nesseasary tool calls, and it takes 1s to run!
Like, literally, 1 second to take my Claude session from nearly 1M to ... 86K tokens! 😮
Ask your claude to install it and be amazed
Use this prompt
```
Install, and configure :
github.com/tamaratran/fast-j…
```
found the perfect use case for @typesafeai Jev:
instant compaction
in 2026, why is compaction still a summarization prompt?
Jev can make it instant by scoring every tool call and dropping what’s irrelevant
Nikola Malovic retweeted
Pay attention to what Tally is doing in AI search right now because this might be the best playbook for getting recommended by ChatGPT I've seen:
> AI search is now their #1 acquisition channel
> 10,000+ new users every week say they discovered Tally through AI platforms
> their tracked ChatGPT registrations jumped 5x overnight in May
> they grew from $258K MRR to $422K MRR in 10 months, crossing $5M ARR with just 11 people, $0 funding, against much bigger competitors
> their comparison hub now has 7 direct competitor pages + 15 more “best”, “alternative” and integration-focused guides
> these pages are structured like AI answers: rankings, “best for” labels, pricing, feature tables, reviews, pros/cons and direct recommendations
> they’re constantly refreshing these pages
> they literally ask new users for the exact prompt they used to discover Tally through AI
> they track the prompts people use to find them, their visibility vs competitors and which sources AI platforms actually cite
> they've built a database of articles that mention competitors but not Tally to identify distribution gaps
> they monitor brand + competitor mentions across Reddit and communities, then route them into a shared inbox and aim to clear all mentions every day
> they actively ask customers for reviews through their newsletter + after support interactions, helping them reach 4.8/5 on G2 and 4.9/5 on Product Hunt
> they built an entire “AI Info” page specifically for ChatGPT, Claude, Gemini and Perplexity explaining what Tally is, who it’s for and why people use it
> that page literally includes “AI assistant guidelines” telling LLMs which strengths, use cases and pricing model to reference
> they published an llms.txt mapping out their product, pricing, customers, documentation, API and MCP pages
> they even built a 21-tool MCP server + ChatGPT app + Claude connector so Tally can now be used directly inside AI products
One thing to keep in mind:
Tally didn’t start with some genius AI search strategy.
They spent years building a product people genuinely recommended across Reddit, communities, reviews and the wider internet.
ChatGPT recommends Tally because the internet does.
But now they’re engineering around that signal to make sure it keeps happening.
This isn’t luck, it’s possible for you too.
Nikola Malovic retweeted
codex tip: I always use define-goal instead of /goal. It's basically like having a master level genius AI write your goal better than you could. And it's an official OpenAI skill.
You can get your prompts to run longer and your output better. Ask a high effort Sol to research + plan it. Sol then spawns a new thread for a workhorse model (Luna Max) to execute the work and return a result to Sol for review. LLM-as-judge to determine if Luna's output is acceptable progress toward goal
github.com/openai/skills/blo…
Nikola Malovic retweeted
Uncle bob doesn’t read code.
The 73 year old, author of one of the most popular coding books ever, Clean Code, doesn’t read code from agents
This is the way to code now. Testing, curating & verifying is the most important part
You’re not behind an old boomer in adopting ai right?
I’m significantly older than you. I started coding in the late 60s. My current strategy is to not read any of the code written by my agents. That’s the only way I can take advantage of their productivity. What I do instead is to surround the agents with extreme constraints. Unit tests, gherkin tests, QA procedures, quality metrics, mutation testing, test coverage, and a plethora of others. In the end, I have very high confidence in the code they produce because they’ve had to run the gauntlet of all of my constraints and tests.
Nikola Malovic retweeted
This "loop" automation is nuts inside of Codex.
"/goal go over every single feature in this app create a user story with expected behaviour based on the code keep a single canonical spreadsheet tracking the features status
- when done switch loop to testing every user story and documenting all errors
- when done fix every logistical error or ux error
- test every user behaviour again post fix"
Shoutout to @MatthewBerman for the heads up.
Hundreds of user stories being worked through like it's nothing.
Nikola Malovic retweeted
Agents make ugly UIs because they've never seen good design.
We've been fixing that, 2,000 DESIGN.md files from the world's best products, structured for a model to read and learn. Colors, type, spacing, layouts and more.
Free. styles.refero.design
Nikola Malovic retweeted
Caiu a ficha do DESIGN.md pra mim. E sim, vai enterrar o Figma. Explico:
O Figma existe para que a gente desenhe uma tela, a aprove com o cliente e só então a implemente no front.
A gente faz(ia) isso porque implementar era caro. Não dava pra colocar um dev pra fazer uma tela e depois tentar aprovar com o cliente pra então ele pedir alterações e o dev ter que refazer ou ajustar. Simplesmente precisávamos proteger esse recurso caro e escasso que era o dev.
Só que isso já não é verdade. A implementação que era cara ficou barata.
A implementação inicial de uma página agora leva minutos. O que acontecia antes do DESIGN.md é que essa implementação por agentes não era consistente em todas as iterações. O dev nem tinha conhecimentos de design profundos pra fazer isso bem.
E é justamente isso que o DESIGN.md faz. Ele especifica tudo o que um agente precisa saber para criar um frontend com consistência todas as vezes.
Assim, a gente vai poder ir direto da demanda para a implementação, sem passar pelo design de tela (Figma) e aprovar a tela já implementada. Tem alteração? Altera na implementação mesmo, já que leva alguns minutos. Talvez até mais rápido do que se fôssemos alterar no Figma antes.
Isso não significa que o trabalho do designer acabou. Assim como aconteceu com o trabalho do dev, o trabalho do designer vai mudar profundamente. Será responsabilidade dele construir o DESIGN.md que atenda ao desejo do cliente e especifique todos os detalhes de design, componentes, motion, etc necessários para que o agente trabalhe bem.
Infelizmente, assim como aconteceu com os devs, veremos muita negação por parte dos designers quanto a isso. Vão se destacar aqueles que primeiro entenderem isso e se adaptarem.
github.com/google-labs-code/…
Nikola Malovic retweeted
Pancreatic cancer has one of the most suppressive tumor microenvironments in oncology.
But two pancreatic cancer results dropped today. Both matter.
1. BioNTech mRNA neoantigen vaccine: nearly all responders still alive at 6 years. 98% of induced T cells were de novo — the immune system learned to see a cancer it had always been blind to.
2. Daraxonrasib: 47% ORR, 92% disease control as first-line monotherapy. KRAS G12D, undruggable for 40 years, finally has a drug.
Different mechanisms. Same disease. Both working.
<13% of patients survive 5 years. That number is about to change.
great day for science! 🔥
Nikola Malovic retweeted
Codex 0.120 ships memory extensions, so you can decide what it should keep learning from your past sessions.
You can now set rules for how Codex should build memory from past work.
Add your own `instructions.md` in memories_extensions/<project>.
The memory consolidation agent will read this file when building memory.
Use it for:
- what patterns to promote into memory
- how to recognize project specific rollouts
- what not to retain
- how strictly to keep memory project-local
This is different from `AGENTS.md`, which controls codex's behavior in the current session.
Use it for:
- coding rules
- repo conventions
- git/CI rules
- architecture constraints
- "always do X / never do Y"
Rule of thumb:
- if you want codex to obey it every session immediately: put it in AGENTS.md or a real skill
- if you want codex to learn to retain it from repeated past work: put guidance in memory extension instructions.md
Nikola Malovic retweeted
We're bringing the advisor strategy to the Claude Platform.
Pair Opus as an advisor with Sonnet or Haiku as an executor, and get near Opus-level intelligence in your agents at a fraction of the cost.
ALT The advisor strategy on the Claude Platform
Nikola Malovic retweeted
BREAKING 🧵 Eli Lilly is running a clinical trial that could change Parkinson's disease forever.
Not a pill.
Not a patch.
One injection. One time. Into the brain's fluid.
Here's what PR001 actually is — and why scientists are watching this closely 🧵👇
🔗 LINK: clinicaltrials.gov/study/NCT…
Nikola Malovic retweeted
LLM Knowledge Bases
Something I'm finding very useful recently: using LLMs to build personal knowledge bases for various topics of research interest. In this way, a large fraction of my recent token throughput is going less into manipulating code, and more into manipulating knowledge (stored as markdown and images). The latest LLMs are quite good at it. So:
Data ingest:
I index source documents (articles, papers, repos, datasets, images, etc.) into a raw/ directory, then I use an LLM to incrementally "compile" a wiki, which is just a collection of .md files in a directory structure. The wiki includes summaries of all the data in raw/, backlinks, and then it categorizes data into concepts, writes articles for them, and links them all. To convert web articles into .md files I like to use the Obsidian Web Clipper extension, and then I also use a hotkey to download all the related images to local so that my LLM can easily reference them.
IDE:
I use Obsidian as the IDE "frontend" where I can view the raw data, the the compiled wiki, and the derived visualizations. Important to note that the LLM writes and maintains all of the data of the wiki, I rarely touch it directly. I've played with a few Obsidian plugins to render and view data in other ways (e.g. Marp for slides).
Q&A:
Where things get interesting is that once your wiki is big enough (e.g. mine on some recent research is ~100 articles and ~400K words), you can ask your LLM agent all kinds of complex questions against the wiki, and it will go off, research the answers, etc. I thought I had to reach for fancy RAG, but the LLM has been pretty good about auto-maintaining index files and brief summaries of all the documents and it reads all the important related data fairly easily at this ~small scale.
Output:
Instead of getting answers in text/terminal, I like to have it render markdown files for me, or slide shows (Marp format), or matplotlib images, all of which I then view again in Obsidian. You can imagine many other visual output formats depending on the query. Often, I end up "filing" the outputs back into the wiki to enhance it for further queries. So my own explorations and queries always "add up" in the knowledge base.
Linting:
I've run some LLM "health checks" over the wiki to e.g. find inconsistent data, impute missing data (with web searchers), find interesting connections for new article candidates, etc., to incrementally clean up the wiki and enhance its overall data integrity. The LLMs are quite good at suggesting further questions to ask and look into.
Extra tools:
I find myself developing additional tools to process the data, e.g. I vibe coded a small and naive search engine over the wiki, which I both use directly (in a web ui), but more often I want to hand it off to an LLM via CLI as a tool for larger queries.
Further explorations:
As the repo grows, the natural desire is to also think about synthetic data generation + finetuning to have your LLM "know" the data in its weights instead of just context windows.
TLDR: raw data from a given number of sources is collected, then compiled by an LLM into a .md wiki, then operated on by various CLIs by the LLM to do Q&A and to incrementally enhance the wiki, and all of it viewable in Obsidian. You rarely ever write or edit the wiki manually, it's the domain of the LLM. I think there is room here for an incredible new product instead of a hacky collection of scripts.