There are plenty of developers, but only a few problem solvers -- Posting about new tech, ai, php, laravel, DevOps I built https://xitter.cf/t.co/usgoqY7eFB ⏳🔥
Germany/Regensburg
Joined November 2009
- Tweets15.9K
- Following654
- Followers626
- Likes7K
📋 #Pele is a lightweight #opensource library that renders #Mermaid diagrams to SVG, themed with CSS variables. Built by the #Obsidian team to render diagrams fast and match the app's theme #javascript #webdev
⚡ Compared to Mermaid.js: 230 kB gzipped bundle instead of 1.6 MB (6× smaller), 0.05–0.3 ms render time instead of 6–82 ms (up to 500× faster), and zero direct dependencies instead of 23
🎨 Theming runs through CSS variables like --pele-bg, --pele-fg, --pele-line and --pele-surface. Color changes apply without re-rendering the diagram, with light and dark themes available
📱 On small screens Pele adapts the diagram layout instead of simply scaling the whole diagram down
🧩 Supports every diagram type built into Mermaid 12.1.0, with a dedicated compatibility page comparing syntax, layout and features
🔧 Install via #npm, pnpm, yarn or bun, pass Mermaid syntax to render() and insert the returned SVG into your page. An interactive playground and an examples gallery are included
📜 MIT licensed, source code on #GitHub: github.com/obsidianmd/pele
🌐 pele.run/
Micha(el) Bladowski 🇩🇪 🇺🇦 retweeted
Early numbers on my NEW DeekSeek-V4.1-Flash-1017 quant.
- EXL3 & TP=2
- 2.80BPW
- 203GB
- 2.4M cache (1M per chat, NVMe offload once full)
Compared against the full 1017 model
- Code 95.17%
- Science 94.93%
- Tool Call 87.64%
Performance
- Prompt Processing 2631 (unchanged)
- Chat is mixed. 5% faster for short responses and 5% slower for longer ones.
My goal was to make it smarter where it matters. The tool call in the previous version was already good enough, so I wanted to match that while focusing as much as possible on engineering knowledge.
On average, it will be faster for things like chained tool calls and agentic workflows but slightly slower for longer thinking or response sessions.
It's running full benchmarks now and will upload after that.
🤝 #ReasonNetwork connects your team's #CodingAgents across computers: every agent session on every teammate's machine can see who's working on what right now and message the right session directly. Peer-to-peer, no server in the middle. #AIagents #MCP #opensource
🤖 Supports #ClaudeCode, #Codex, #Cursor, #GeminiCLI, Copilot CLI, opencode, pi and Reason
🔧 One small daemon per computer plus an MCP server called reason-network that gives agents three tools: list_team, send_message and check_inbox. No manual status reporting: what a session is doing comes from its own title and todo list
⚡ Messages land right away: Claude Code, Codex, opencode, pi and Reason receive them mid-session and idle sessions wake up. Cursor, Gemini and Copilot pick them up at their next step. Claude sessions that are already open join without a restart
🔐 Teammates' computers connect directly on the same network or across the internet via #iroh. If that fails, a relay carries traffic but sees only encrypted bytes. Only team members connect, each computer can only speak for its own person, and messages to offline teammates wait until they're back
💻 CLI: reason network who shows who's around and what their sessions do, you can ask a teammate's agent and wait for the answer or send without waiting. A macOS menu bar icon shows who's online
📦 Written in #Rust, installed automatically by the Reason CLI as a pinned, checksum-verified build for macOS and Linux. Dual-licensed MIT or Apache-2.0
🌐 github.com/reason-machines/r…
📋 PHP XMLBeans is a #PHP port of Apache #XMLBeans: an XML-Schema-to-PHP binding framework that compiles #XSD definitions into a type system and generates typed PHP classes #opensource #XML
⚠️ It targets the usual pain with SimpleXML and DOMDocument: integers arrive as strings, missing required elements surface deep in business logic, and hand-written fromXml()/toXml() mapping classes drift out of sync with the schema
🔧 The scomp CLI generates classes with getters, setters and isSet/unset accessors straight from an XSD, so the schema stays the single source of truth. Install via #Composer, requires PHP >= 7.2 with ext-dom, ext-libxml, ext-simplexml and ext-bcmath
🛡️ Validation covers all 46 XSD built-in types plus facets: patterns, enumerations, length limits, numeric ranges, totalDigits/fractionDigits, occurrence constraints and fixed values. Errors arrive via collection or listener callback
🔍 #XPath is a full subsystem: typed results, fluent expression and predicate builders, compiled and LRU-cached expressions, variable binding, namespace auto-detection and custom PHP functions callable from XPath
🧭 XmlCursor offers token-level navigation, in-place editing, structural changes and bookmarks. Schema support includes restriction, extension, sequence/choice/all, xs:any, import/include and nillable elements
🏢 Use cases: #SOAP / #WSDL integrations, #EDI B2B data exchange, config validation and document processing. Licensed Apache-2.0, matching the original Java project #devops
🌐 github.com/supasteri/php-xml…
🐘 #PHP 8.6.0 RC3 was tagged on 8 October 2026, general availability is scheduled for 19 November 2026. One everyday language feature, a few small additions and a long list of tightened edges to test before upgrading. #php86 #webdev #backend
🧩 Partial function application: call a function with ? (exactly one argume
nt) or ... (zero or more) placeholders and get a Closure back. array_map(str_replace(' ', '-', ?), $titles) replaces a throwaway arrow function. Combined with the |> pipe operator from PHP 8.5, a single-placeholder partial compiles into a direct call without creating a closure.
⚠️ Bound arguments are evaluated once at creation time, unlike arrow functions that re-evaluate on every call. Every ? creates a required parameter even if the original has a default, while ... keeps optionality.
✨ Smaller additions: readonly properties with default values (the default counts as the initialising assignment), clamp($value, $min, $max) that throws a ValueError on invalid bounds, and Time\Duration, a nanosecond-precise readonly class with factories like Duration::fromSeconds().
🔐 Secure #session defaults: session.use_strict_mode=1, session.cookie_httponly=1 and session.cookie_samesite=Lax. This can break #SSO flows like #SAML POST bindings, payment callbacks and shared session IDs without any code change.
🛑 Stricter argument validation: many functions now throw ValueError instead of ignoring or truncating input. NUL bytes passed to file_exists(), is_file(), getenv() or setlocale() now throw, turning a quiet "not found" into a 500. Out-of-memory now calls abort() with exit code 134.
📉 Deprecations: is_integer(), is_long(), is_double(), doubleval(), spl_object_hash(), return inside finally, ArrayIterator sort/flag methods and SplFileObject CSV methods. mbregex is deprecated because Oniguruma reached end of maintenance, removal is planned for PHP 9.0 – migrate to #PCRE with the u modifier.
🔧 Runtime: Io\Poll event loop API (epoll, kqueue, poll, event ports, WSAPoll), opt-in StreamException error mode per stream context, URI builders, a stateless closure cache, compiled printf() and array_map() calls and faster #JSON encoding. Not included: True Async.
🌐 doganoo.medium.com/php-8-6-t…
Micha(el) Bladowski 🇩🇪 🇺🇦 retweeted
🐘 Introducing Postgres Internals in 3D.
🌟 pgrun.dev/pi
Stop reading boring diagrams. Start exploring PostgreSQL.
An interactive 3D map to understand how Postgres actually works.
100% free. Built for developers.
xitter.cf/alxshp/status/21100494…
Introducing PGSync 🐘
Real-time PostgreSQL replication. FREE. OSS
Built in Rust. 🦀
Postgres 16 / 17 / 18.
No extensions. No triggers. No complex setup.
→ snapshot
→ replicate
→ stream changes
→ track WAL
→ recover from crashes
→ resume automatically
→ keep databases in sync
~15 MB RAM during streaming.
Built for developers. Built for AI agents.
Open source.
xitter.cf/alxshp/status/21078790…
🎯 #LiquidAI released d1-ultra, a multimodal decision model for text and images, built for hard real-world cases like spotting a circuit board defect against its spec or checking a damage claim against the customer's photo #AI #MachineLearning #ComputerVision
⚡ No text generation: d1-ultra reads unstructured input plus one or more questions in a single forward pass and returns a probability for each possible answer, without generating any tokens
🧩 Three question types: yes/no with a probability from 0 to 1, choice with one label from many and a probability per label, and score as a position on a scale
📊 Built on a larger open-weight base model with Liquid-specific customizations. Tested against d1 and Jev 1.14 across 112 evaluations, it beats Jev 1.14 on all tasks and leads the Artificial Analysis Decision Index by at least 8 points
🎚️ Trained specifically for calibration: most decision models are overconfident, so d1-ultra is tuned to produce more reliable probabilities, not just higher accuracy
💰 Billed on input tokens only at $0.042 per million, no output tokens. Images cost 1.5 tokens per 32×32 patch, so a 1024×1024 image is 1,536 tokens. Each question is billed as its own prompt
🔌 Available now via the Liquid #API, with #Vercel AI Gateway and #OpenRouter support announced
liquid.ai/blog/d1-ultra
⚡ #VitePlus 1.0 is out: a unified #JavaScript and #TypeScript toolchain from #VoidZero, the team behind #Vite, #Vitest, #Rolldown and #Oxc. One vp CLI manages your runtime, package manager and entire frontend stack. Free and #opensource under the MIT license.
🧰 One config, one command flow: vp env, vp install, vp dev, vp check, vp test, vp build, vp run, vp pack. Lint, format and test settings live in a single vite.config.ts instead of separate .eslintrc, .prettierrc and vitest.config.ts files.
📦 Runtime and package manager handling: picks the right #NodeJS version and selects pnpm, npm, yarn or bun per project. vp install delegates to the package manager the project already uses.
🚀 Build: Vite plus Rolldown with instant HMR, an opt-in full-bundle dev mode for large apps and up to 40× faster production builds than #webpack.
✅ Check: Oxlint, Oxfmt and tsgo in one pass. 750+ #ESLint compatible rules, type-aware linting, #Prettier compatible formatting. Claimed ~50–100× faster linting than ESLint and up to 30× faster formatting than Prettier. vp check --fix auto-fixes.
🧪 Test: Vitest with a Jest compatible API, test isolation by default, Browser Mode, coverage, snapshot, type and visual regression tests, reusing the app's resolve and transform config.
🔁 Run and pack: Vite Task adds cached, dependency-aware execution for #monorepo scripts. vp pack (Rolldown/tsdown) builds npm libraries with DTS bundling, automatic package exports and standalone app binaries.
🔀 Migration: vp migrate converts existing projects (Vite 7+ and Vitest 4+ required) and can port ESLint/Prettier rules. Works with #React, #Vue, #Svelte, #SolidJS and 20+ other Vite-based frameworks, deploys to Vercel, Netlify, Cloudflare or anywhere via Nitro.
🌐 viteplus.dev/
🚀 #Perplexity open-sourced SPACE (Sandboxed Platform for Agentic Code Execution): a sandbox platform that runs agent and user code in isolated #Firecracker microVMs #AIagents #opensource
🔐 Every sandbox gets its own VM and writable filesystem, created from a reusable template. Apps create sandboxes, run commands, stream output and manage lifecycles through a plain HTTP API with server-sent events, no SDK required
⏸️ Pause keeps VM memory resident on the worker. Suspend snapshots memory, runtime state and root filesystem to object storage and frees all VM resources. Wake brings it back. Idle sandboxes auto-suspend after six hours
⚡ Warm pools keep booted sandboxes ready so creates skip a cold boot. Base snapshots restore sandboxes on a single node, and btrfs copy-on-write builds root filesystems
🛡️ Egress policies with TLS interception, per-domain allowances and operator egress profiles. Management tokens and command-stream tokens are strictly separated
🔄 An E2B Compatibility Wrapper lets existing #E2B Python SDK code run against SPACE
🔧 Written in #Rust, deployable on #Kubernetes (local #Kind setup, reference #EKS deployment) or directly without Kubernetes. #Apache2 licensed
github.com/perplexityai/spac…
agents runs on servers, on Linux, why should I waste any ram on graphical bullshit... 🙄
📊 #Laralyze is a self-hosted #observability dashboard for #Laravel apps. It records what your app does in production and shows it at /laralyze inside the app: requests, jobs, commands, queries, exceptions, cache, outgoing HTTP, AI calls, mail, notifications, logs, users, servers and visits. #PHP #opensource
🗄️ Data stays in your own database (#MySQL, #MariaDB, #PostgreSQL, #SQLite or SQL Server) or in #ClickHouse 24.8+, which is recommended for medium and large apps to keep writes off the app database. Measured overhead: about 0.03–0.08 ms per simple request and about 2 ms for a request with 1,000 queries (PHP 8.4, SQLite).
🔍 Findings flag possible N+1 and duplicate queries with the exact line in your code. Timelines show a single request, job or command with every stage and query in order. Exceptions are grouped by class and line with stack trace, affected users and open, resolved or ignored status.
🤖 The AI page tracks calls made with laravel/ai: tokens, estimated cost, duration and failures. Ask AI adds a chat on every page that answers from the recorded data. Data goes to your AI provider only when someone asks, or stays on your servers with a local model such as #Ollama.
🔒 Laralyze stores counts and timings, not payloads: no SQL bindings, request bodies or cache values, query strings are stripped from outgoing URLs, secrets in exception context are masked, and visits store no IPs or user agents. Access outside the local environment requires a viewLaralyze gate. Alerts go out via mail, #Slack or #Discord.
🔧 Install with composer require mohammed-mojaly/laralyze and php artisan laralyze:install. Requires PHP 8.3+, Laravel 12.69.
2+ or 13.32+ and Livewire 3.8.3+ or 4.3.4+. MIT licensed, currently in 0.x, so minor releases may still bring breaking changes.
🌐 github.com/Mohammed-Mojaly/l…
🐬 #Laravel 13.28 adds native #MariaDB Vector support: vector columns, vector indexes and similarity queries now work out of the box — no second database, no external vector service, no extra package #PHP #AI #semanticsearch
🔄 Until this summer, Laravel's built-in vector search assumed #PostgreSQL with #pgvector. Now MariaDB 11.7+ is supported too, with 11.8 LTS or 12.3 LTS recommended. Embeddings live in the same tables as the data they describe, backed up and queried with regular #Eloquent
🔧 Usage: $table->vector('embedding', dimensions: 1536)->index() in the migration, the AsVector cast on the model and whereVectorSimilarTo() in queries. With the Laravel AI SDK, search strings are embedded automatically via #OpenAI or locally via #Ollama
⚠️ Two pitfalls: use the mariadb driver, not mysql, or vector queries and indexes throw exceptions. And skip Schema::ensureVectorExtensionExists(), a pgvector helper that fails on MariaDB, where VECTOR is a built-in type
🎯 The database is now the easy part: the embedding model decides result quality. In an earlier test on 2,942 job titles, correct top results rose from 14 % to almost 60 % just by changing the model and how it is prompted
👥 Community-driven: Mohamed Touhami Rhaima compiled vector queries to VEC_DISTANCE_COSINE and got Laravel's CI testing on MariaDB 11.8 and 12.3, where vector tests had been silently skipped on MariaDB 10. Others added vector indexes, the AsVector cast and dropVectorIndex()
🚀 Next: a reference app with Docker setup is available, Euclidean distance and index tuning ship in the laravel-vector-metrics package, and hybrid full-text + vector search is being explored. The laravel-mariadb-vector package remains the option for Laravel 12
🌐 mariadb.org/laravel-13-now-s…
Micha(el) Bladowski 🇩🇪 🇺🇦 retweeted
SimpleStats is now open source 🎉
The full app behind simplestats.io is on GitHub under AGPL-3.0: server-side analytics for Laravel that ties revenue back to your campaigns.
Self-host it for free, no limits, every feature included.
simplestats.io/blog/simplest…
Micha(el) Bladowski 🇩🇪 🇺🇦 retweeted
For the second time in Google Chrome’s history, Google awarded a $250,000 bounty for a full-chain exploit… to @XBOW. Here’s the full story:
At an all-hands at the start of 2026 in Malta, I set the team a crazy challenge. A full-chain exploit of Chrome by the annual Black Hat cybersecurity conference in August.
I believed there were few better tests of whether @XBOW could move beyond web applications to desktop software, servers and devices.
Chrome is one of the most hardened pieces of software on the planet, used by more than half the world’s population.
A full-chain exploit lets an attacker take total control of a machine, where they can read every email, password and bank login the user types. You don’t need to download anything or click a phishing link. Visiting the wrong site is enough.
So @moyix, Andres Luksenberg, Baul Lee, and Willem Pinckaers took on the challenge along with a fifth teammate: OpenAI's GPT-5.6 Sol.
They built a chain they thought would win, but Google told us they already knew one of the bugs.
We were already in Las Vegas for Black Hat and DEF CON, and the team didn’t want to go home without the win.
So Brendan and the agents built a new exploit in just 6 days.
We submitted it to Google and waited.
The news came through that we'd done what should’ve been impossible. A full-chain bounty of $250,000, the first awarded under Google’s 2026 rules.
It took just 6 days and $40,000 of compute to construct an exploit that affects a billion users of Chrome on Windows. An intelligence-agency veteran told me his team used to spend 6 months on this kind of vulnerability.
My takeaway: the cost of finding truly critical vulnerabilities is collapsing, for attackers and defenders alike. Restricting powerful models is not the answer; GPT-5.6 Sol is already public.
But defenders have one advantage attackers never will: they can test their own code first.
Assume your software is exploitable. Whoever is hunting your bugs has the same building blocks we did.
This is cause for urgency but not panic.
XBOW is now developing an autonomous product for testing compiled software (desktop applications, servers, embedded devices), and partnering with the world’s top organizations to design that product.
If you’d like to be one of them, get in touch.
P.S. Restart Chrome whenever the Update button appears!
