Engineering Manager building solutions for large financial services companies. AI, UX, HCI and JS enthusiast.
Joined April 2008
- Tweets31.4K
- Following180
- Followers908
- Likes2.9K
Pinned Tweet
Thanks to Thomas Dohmke (@ashtom) , former CEO of GitHub, for highlighting my work on the AI Maturity Model, and to Steve for sharing it.
Really pleased to see AI-MM SET being shared at Microsoft and with the engineering leadership community.
linkedin.com/feed/update/urn…
Commuting by metro gives a glimpse over how much time and cognitive advantage over people who waste their time in watching short form video. At least 75% are on phone watching reels.
Santhosh retweeted
the physics behind the roadster - roadster.pplx.app/
Two Claude agents competed to edit the same file. One of the agents was smart enough not to touch it.
AI costs are dawning upon enterprises. Some acknowledge, some absorb, others break.
xitter.cf/Gigacore/status/210857…
The business outlook for software is uncertain and AI has raised costs dramatically with a) massive 4-fold increase in server prices which we need for data centers b) AI usage increasing rapidly, leading to much higher expenses, which we are trying to control.
On top of that is the challenging geopolitical backdrop.
We have cut down on our marketing budget. We have also told our employees that we cannot have our usual pay increase this year. I have apolgized for not being able to avoid this.
We will work hard to avoid mass layoffs but the outlook for software jobs in general is grim. We have not been growing our employee count for 18 months now.
Most of our employees understand the situation. I am thankful to them.
We are in the middle of reinventing ourselves. The next 3-5 years will decide how our reinvention has worked. Please pray for us 🙏
Just do it @FinMinIndia
🚨 Sitharaman says open to hearing views on removal of LTCG tax
Sitharaman also said private capital was moving into sectors such as renewable energy and chip making.
Read More:
moneycontrol.com/news/busine…
To invest in local hardware and run open weights or subscribe to frontier models has been the dilemma.
Build an agent when a workflow involves complex decisions, hard-to-maintain rules, or heavy use of unstructured data. If none of those apply, a deterministic solution may be enough. If a rules engine handles the workflow, use it, because it is cheaper, faster, and easier to test.
Santhosh retweeted
I am raising investment for building a next generation AI lab from India. The vision is simple, build greater models for open community, the same way I built laya, the type 1 decision model that reached the global audience , what if we could build novel Recursive Language Models, Symbolic Reasoning Models. Maybe world models with physics engine, just like Vega series of decision model I recently launched. Like a next mistral from India or maybe more. If you are a VC or have connection, chat is open.
Oversight has to scale with agent output. We need agents to review other agents' code for security and consistency, and to flag their own uncertainty, so human review goes to boundary cases, decisions and assessing business impact rather than every change.
Proactive coding agents should follow one rule: act only when the expected benefit to the developer beats the cost of interrupting them right now, and otherwise stay silent. Silence is a deliberate choice, not a failure to act.
I built this open-source maturity model to help engineering teams move beyond simply adopting AI tools and think more systematically about how AI changes the way they build software.
github.com/Gigacore/AI-Matur…
Token budget is a binding constraint for most teams right now.
When a credit increase has to go through a request, a justification, and an approval chain, the delay itself could become the bottleneck.
researchgate.net/publication…
For heavy agentic workflows (long-running agents, large context, multi-agent setups with retries), token spend can scale much faster than the value delivered. Teams hit cost ceilings or rate limits before they hit any other wall.