Mohammad Alaggan, Ph.D. (@malaggan) header
Mohammad Alaggan, Ph.D. (@malaggan) profile picture

Principal Engineer at @AWSCloud (CloudFront). Opinions are my own.

Toronto, Canada
Joined October 2009
Photo by Mohammad Alaggan, Ph.D. (@malaggan)Photo by Mohammad Alaggan, Ph.D. (@malaggan)Photo by Mohammad Alaggan, Ph.D. (@malaggan)Photo by Mohammad Alaggan, Ph.D. (@malaggan)Photo by Mohammad Alaggan, Ph.D. (@malaggan)Photo by Mohammad Alaggan, Ph.D. (@malaggan)Photo by Mohammad Alaggan, Ph.D. (@malaggan)Photo by Mohammad Alaggan, Ph.D. (@malaggan)Photo by Mohammad Alaggan, Ph.D. (@malaggan)Photo by Mohammad Alaggan, Ph.D. (@malaggan)
Sounds like today's politicians (also mathematicians vs scribes) "From the Sultan's perspective, the printing press caused nothing but trouble; it educated people and gave them access to dangerous ideas. It's much more about the fact that the Sultan saw it as a danger to his power!"
Insiders of the Islamic World have that nasty habit of blaming the Ulema for everything bad in the Islamic World. It's driven by contemporary politics but it shouldn't be projected back in time! The Ulema played a minor role in the Ottoman ban on the printing press; I would say a bigger role was played by lobbyists from the scribing industry. The biggest question that no one answers is, "What's the perceived problem that the printing press would've solved in 1500 in the Islamic World?!" From the Sultan's perspective, the printing press caused nothing but trouble; it educated people and gave them access to dangerous ideas. It's much more about the fact that the Sultan saw it as a danger to his power!
1
122
I cancelled my Claude subscription, will cancel my ChatGPT subscription as well, and will try t3chat just to use Neon as my default workflow from now on.
Gemini Neon unexpectedly brings me deep deep happiness. I never thought I’d feel that way from using a model, not in 2026. But it’s here! It’s so creative and easy to chat with. In my mind it’s in many ways better than Fable. It’s such a blessing.
2
251
Mohammad Alaggan, Ph.D. retweeted
People complain Claude didn't "invent new branches of mathematics" in solving the Riemann Hypothesis. But Galois developed group theory only after Abel had already proven the quintic’s insolvability. Partly to better explain that proof. Mathematical theories are the framework we build to better encapsulate the details of a complicated proof. Not a necessity.
Mathematics via infinite monkeys theorem I mean... a proof at 250,000 pages, what are we even doing at this point
22
26
7
360
24,150
Neon 4 is absolutely amazing
quick TL;DR on Gemini 4 Neon by Theo > expected a bad medium sized model from google > but it turned out to be the best coding model > google understated gemini 4 neon > carbon is garbage and neon is great > neon took 10 hours for a task which opus and astra couldn't finish in 10+ days > neon is more coherent than claude and gpt > neon writes code he's more likely to merge > neon still uses em dashes > neon is not so good at 3d > antigravity app is garbage (for now) > new frontier is not capability, it is coherency > best of muse + best of astra = gemini 4 neon > it didn't trigger his custom skill (only issue) > it could have been named gemini 5 > it is theo's favorite model right now NOT A PAID SHILL OF GOOGLE
7
391
Gimini 4 Neon (the smartest model I've seen so far) is losing its mind over a proof written by Astra 6.2 over 2 weeks.
1
138
Mohammad Alaggan, Ph.D. retweeted
We audited Masullo’s manuscript The Music of the Primes, which claims an unconditional proof of the Generalized Riemann Hypothesis. In its current form, the paper's argument has fundamental problems in several distinct parts of the strategy. 🧵 1/
This quoted post is unavailable.
19
54
18
479
129,366
Mohammad Alaggan, Ph.D. retweeted
We have a full confirmation from Andrew Sutherland who first managed to finish the nanoda kernel run and comparator in paranoid mode. This is now basically saying that the Riemann Hypothesis is 99,99999% true. It's almost impossible that any glitch has been deliberately implemented into the code and that anything is missing from the proof artifact (no sorries, no additional axioms, no weird bugs).
We don't yet have full confirmation from the paranoid-mode comparator test, but we are close to seeing a formal artifact generated in Lean that provides a proof of the famous Riemann Hypothesis. This is such a historic moment. The code was generated by @__alpoge__ using Claude and possibly other software. There is still some uncertainty, but we may be on the verge of a historic announcement. For the first time in history, we may have such a magnificent proof that we DON'T fully understand yet, but that is probably correct.
68
356
71
2,788
361,053
Mohammad Alaggan, Ph.D. retweeted
Replying to @thorstenball
It's often faster to prompt an agent and make a change, than to think about what the right thing to do in the first place is.
I recently got a small, possibly vibed PR on my open source project. It (cl)aimed to fix a long standing open issue. The reasoning seemed plausible, and my review agent had no complaints, so I was tempted to just LGTM it. After all it was late in the evening, and who reads code these days anyway. Still, I took a peek at the diff, and it seemed plausible enough, a simple `if` plus some tests to cover the case. But then I started thinking about these changes and I became a bit puzzled, so I decided to relay my confusion to the author. When I got the answer, it became clear that the PR was fixing the wrong thing. The issue was an incorrect usage in the client code. And the deeper problem was that the documentation wasn't quite clear about it. Also, the library could detect this situation and report an error. I could have asked the author to forward the request to their agent, but seeing how the holiday season is approaching I decided to burn my own tokens instead. I asked my agent (using Sol 6.2) to make the changes, took a look at the diff, and the code felt a bit too elaborate. Basically, we have to verify that in a small list the element `:boundary` comes before the elements `:elixir` and `:app`. To do this, the agent combined `Enum.each` with `Enum.find_index` and a `cond`. When I pointed that it could simply filter those three elements and compare the resulting list to `[:boundary, :elixir, :app]`, it concurred and applied my suggestion, but for good measure added a pinch of signature slop. I sighed and did something unspeakable - I edited the code manually 🙈. For good measure I also deslopified docs & tests. As I was sipping some well-earned elixir, I reflected on the symbiosis of a human and an agent. Clearly together we came up with an improvement. Provided with an exception stack trace an agent was able to quickly come up with a failing case. I wouldn't be able to think of this misuse example myself. But it also produced an inappropriate solution, and later generated an overcomplicated code. I guess it's a skills issue. Skills such as reading, thinking, and understanding 🙃
1
1
56
Mohammad Alaggan, Ph.D. retweeted
Formal verification isn't finding bugs, it's proving there are none. But what if you attempt to formally verify code that has a bug? The second post on our x86 formal verification research project covers bugs we found in glibc, and how failing proofs can lead right to them.
10
41
4
302
18,053
Mohammad Alaggan, Ph.D. retweeted
Each word you say is planned by a small patch of brain above your left ear, about a quarter of a second before you say it. It sits in a temporal lobe region textbooks describe as only for listening. Our paper today in @ScienceMagazine: science.org/doi/10.1126/scie… @UCSF @NeurosurgUCSF
197
1,589
174
9,930
1,212,986
Mohammad Alaggan, Ph.D. retweeted
wow. this is pretty wild. first, Eddie Chang is now on X (automatic follow!) second, the Chang Lab just solved a longstanding neuroscience puzzle. using single-neuron neuropixels recordings + high-density ecog + intraoperative cortical stimulation, they showed in @ScienceMagazine that a small patch within the brain’s classical “language comprehension” region (wernicke’s area) is actually critical for generating human speech, and therefore a promising new target for speech BCIs textbook-changing work!
39
366
9
4,131
250,419
Mohammad Alaggan, Ph.D. retweeted
Introducing Frostbite: a new class of CPU built around interaction nets. Given only interaction net theory, our agents co-designed the ISA and chip architecture from scratch, running large swarms to design, verify, and optimize the CPU down to the final GDS. On our first set of target workloads, Frostbite is 4.6x more performant per watt than modern traditional CPUs at similar area.
106
328
135
3,099
298,800
Mohammad Alaggan, Ph.D. retweeted
AWS just published an absolute banger of a paper at this year’s SOSP on their use of formal methods to validate S3’s new distributed metadata index. I’m yet to read it, but it builds on their 2021 paper on verifying S3’s Shardstore with formal methods (which I absolutely loved!)
15
45
524
23,263
Mohammad Alaggan, Ph.D. retweeted
“Everything fails, all the time." So we stopped waiting for failover. Multipath Resilience is now GA: identical packets across physically diverse AWS backbone paths at once. First one in wins. One flag to turn on. 99.999% SLA. Live today with @stripe and @boostmobile. go.aws/4kmJwlo
7
16
8
194
32,820
Category theory
Learning Abstract Algebra makes you realize there are only like a handful of algorithms.
1
4
366
Mohammad Alaggan, Ph.D. retweeted
We just shipped AWS Emulator, which gives coding agents a local copy of AWS to test against. Early in my career, we built test environments like this by hand, one system at a time. Now, agents are writing code far faster than any team can check it in production. Matt Goldberg, a Senior Principal Engineer on our Developer Tools team, wanted to know whether agents could build and test those environments themselves. He gave @kirodotdev a single prompt, and about 15 minutes later it came back with a first version of an SQS emulator. Turing that into a robust solution took much more work Here’s what I think this means for engineers today: allthingsdistributed.com/202…
109
259
106
1,804
184,190
Mohammad Alaggan, Ph.D. retweeted
I CANNOT believe the report: it proves that eMudhra is a clown show, and was never fit to be a CA: The point of a HSM (hardware security module) is that the private key NEVER leaves it. Guess what eMudhra did: it turned off FIPS mode (WTF?!) and copied out the private key 😱
118
290
27
4,955
297,228
لما نزلت مصر بعد الدكتوراه عشان ادرس في الجامعة كان موقفي إني ادرس في الجامعة الحكومية المجانية افضل من اني ادرس في جامعة خاصة، عشان ادعم فرص الناس في الترقي الاجتماعي. لما جالي عرض (بدون سعي مني) ادرس «تعمية» في جامعة بفلوس مرضيتش الا اني ادرس نفس المادة في الجامعة الحكومية
الناس مش مدركة فعلًا دور التعليم المجاني في الترقي المجتمعي و أنه كان بيدي للفقير المعدم "فرصة". كشخص جاي من أسرة فقيرة من قرية في قلب الدلتا..مكانش فيه أي دكتور أو مهندس في عيلتي أو معارفي..والنهاردة شغال في أكبر مستشفى مش بس في أمريكا لكن في العالم..فأنا مكنش هيبقى عندي أي فرصة من غير التعليم المجاني اللي حصلت عليه "ببلاش". كان اللي بياخد درس خصوصي بيتبص له على أنه "ضعيف ومحتاج مساعدة زيادة" لأن المدرسة و المدرسين كانوا أكتر من كافيين. أنا اتعلمت يعني إيه ثقافة ومكتبة في مدرسة حكومي في قرية. اتعلمت يعني توسيع مدارك من أمينة مكتبة في حصص المكتبة بتاع المدرسة. مكانش عندي انترنت لحد أولى جامعة (وأعتقد ده كان له تأثير إيجابي أكتر من كونه سلبي) لكن كان عندي فرصة و كان ادامي مصادر و ناس مثقفة ومتعلمة تساعدني في قرية في أقصى غياهب الدلتا. اتعلمت و اتأسست تأسيس هدفع عشرات آلاف الدولارات علشان ابني يحصل على شيء قريب منه دلوقتي في أمريكا. للأسف ده مش موجود دلوقتي للأطفال الصغيرين. و للسبب ده ففعلا أي حد بيدمر أو يحد من فرص تعليم فقراءنا هو public enemy number 1.
2
37
2,213
Mohammad Alaggan, Ph.D. retweeted
we're building a fully native disassembler that doesn't suck
117
98
13
2,297
87,610